INFORMS Philadelphia 1999 Complete Session Listing

Medical Decision Making

SA01.1 Serendipity in Diagnostic Imaging: Magnetic Resonance Imaging of the Breast

• William F. Lawrence; Georgetown University Medical Center, Lombardi Cancer Ctr., 2233 Wisconsin Ave., Ste. 430, Washington, DC 20007; lawrencw@gunet.georgetown.edu
• Wenchi Liang; Georgetown University Medical Center, Lombardi Cancer Ctr., 2233 Wisconsin Ave., Ste. 430, Washington, DC 20007;
• Jeanne S. Mandelblatt; ;
• Karen F. Gold; ;
• Matthew Freedman; ;
• Susan M. Ascher; ;
• Bruce J. Trock; ;
• Polun Chang; ;

MRI of the breast has been proposed for evaluation of suspicious lesions noted on mammogram or clinical breast examination. In a preliminary assessment of this new technology, we use decision modeling to determine the probability of cancer in a lesion found by MRI but not by prior diagnostic work.

SA01.2 Time Tradeoff Evaluation of Temporary Health States in Radiology

• J. Shannon Swan; University of Wisconsin, Dept. of Radiology E3/311, 600 Highland Ave., Madison, WI 53792; jsswan@facstaff.wisc.edu
• Dennis G. Fryback; University of Wisconsin, Dept. of Preventive Medicine, 610 Walnut St., Madison, WI 53705; dfryback@facstaff.wisc.edu
• William F. Lawrence; Georgetown University Medical Center, Lombardi Cancer Ctr., 2233 Wisconsin Ave., Ste. 430, Washington, DC 20007; lawrencw@gunet.georgetown.edu
• Francois Sainfort; University of Wisconsin, Dept. of IE, 1513 University Ave., Madison, WI 53706; sainfort@engr.wisc.edu

We present a new method for assessing utility for diagnostic tests with a modified time-tradeoffs model. The method compares the patient's desire to have a certain test and have treatment now, vs. waiting for results and treatment after an 'ideal' test, with no risk or pain.

SA01.3 What's Really Important? Agreement between Prostate Cancer Patients & their Clinicians

• Gretchen B. Chapman; Rutgers University, Psychology Dept., Busch Campus, 152 Frelinghuysen Rd., Piscataway, NJ 08854-8020; gbc@rci.rutgers.edu,, http://www.rci.rutgers.edu/~gbc
• Arthur S. Elstein; University of Illinois, Dept. of Medical Education, 808 South Wood St., 9th Fl., Chicago, IL 60612-7309; aelstein@uic.edu
• Charles D. Bennett; VA Chicago Health Care System, Lakeside Div., Med. Sci. Bldg., 400 East Ontario St., Ste. 205, Chicago, IL 60611;

Eighty-three patients with newly diagnosed prostate cancer and their clinicians gave time-tradeoff evaluations of 3 multi-attribute health states and the patient's own health state. They also provided attribute importance weights. Patient-clinician agreement was quite low for utilities and weights, but moderate for evaluations of the patient's current health state.

SA01.4 Cost-Effectiveness of Alternative Therapies for Type 1 Diabetes Mellitus

• Francois Sainfort; University of Wisconsin, Dept. of IE, 1513 University Ave., Madison, WI 53706; sainfort@engr.wisc.edu
• Russell H. Tomar; University of Wisconsin, Dept. of Pathology & Lab. Med., 600 Highland Ave., Madison, WI 53792; rh.tomar@hosp.wisc.edu

A simulation model of Type 1 diabetes progression, impact on quality of life and associated medical costs is used to analyze the relative cost-effectiveness of alternative treatment strategies. We report the application of the model, the results and their sensitivity to changes in modeling parameters.

Issues in Network Optimization

SA02.1 A Faster Algorithm for Abstract Maximum Flow

• S. Thomas McCormick; University of British Columbia, Fac. of Commerce & Bus. Admin., Vancouver, BC, V6T 1Z2 , Canada; stmv@adk.commerce.ubc.ca
• James B. Orlin; MIT, Sloan Sch. of Mgmt., E40-147, Cambridge, MA 02139; jorlin@mit.edu

Ford & Fulkerson's 1956 max flow/min cut paper formulated max flow in terms of flows on paths, which Hoffman pointed out still works in an abstract setting. A first constructive proof of this result was recently developed; we show how to speed that algorithm by choosing our augmentations more carefully and taking advantage of the near-network structure this creates.

SA02.2 Implementating Generalized Shortest Paths Algorithms

• Jeffrey D. Oldham; Stanford University, Dept. of Comp. Sci., Gates 4B, Stanford, CA 94305-9045; oldham@cs.stanford.edu

We implement an algorithm for the generalized shortest paths problem. It guesses the solution's cost and uses a Bellman-Ford algorithm to verify whether the guess is too large, correct, or too small. For some instances, our implementation is asymptotically faster than CPLEX. It also differentiates between different negative-cost detection schemes.

SA02.3 The Gale-Shapley Stable Marriage Problem Revisited: Strategic Issues & Applications

• Jay Sethuraman; MIT, 77 Massachusetts Ave., E40-130, Cambridge, MA 02139-3407; jayc@mit.edu
• Chung Piaw Teo; National University of Singapore, 10 Kent Ridge Crescent, Dept. of Decision Sciences, Singapore, 119260 , Singapore; fbateocp@nus.edu.sg
• Wee-Peng Tan; National University of Singapore, , , Singapore;

We consider some strategic issues in the stable marriage problem in which each player is required to submit complete preference lists. We describe an optimal cheating strategy for the players under this model.

Information, Incentives & Coordination

SA03.1 The Option Value of Subcontracting & Outsourcing in Supply Chains

• Jan A. Van Mieghem; Northwestern University, Kellogg Grad. Sch. of Mgmt., 2001 Sheridan Rd., Evanston, IL 60208-2001;

We value the option of subcontracting to improve financial performance and system coordination. We analyze and present outsourcing conditions for 3 contract types: price-only contracts; incomplete contracts and state-dependent price-only and incomplete contracts for which we show an equivalence result. We find that sometimes firms may be better off leaving some contract parameters unspecified ex-ante and agreeng to negotiate ex-post.

SA03.2 Analysis of Capacity Reservation & Spot Bidding under Horizontal Competition

• Sridhar Tayur; Carnegie Mellon University, Grad. Sch. of Indust. Admin., Pittsburgh, PA 15213; stayur@grobner.gsia.cmu.edu
• Pinar Keskinocak; IBM, TJ Watson Research Ctr., Rte. 134, Yorktown Heights, NY 10598; pinar@watson.ibm.com
• Feryal Erhun; ;

An emerging practice that is possible due to Internet-enabled operations, i.e., powder metals industry in the automotive chain, is modeled as a supply chain where 2 outsources may reserve capacity before uncertainty resolution from a common supplier and then bid on spot after while also competing in the same end market.

SA03.3 Order Volatility & Supply Chain Costs

• Rungson Samroengraja; Columbia University, Grad. Sch. of Bus., 3022 Broadway, Uris Hall, New York, NY 10027-6902;
• Fangruo Chen; Columbia University, Grad. Sch. of Bus., 3022 Broadway, Uris Hall, New York, NY 10027-6902; fchen@research.gsb.columbia.edu

Many industry studies show that there is a great potential for cost savings in the area of supply chain management, and many people blame the bullwhip effect as a major source of supply chain inefficiencies. We use a model to show that a replenishment strategy that reduces order volatility does not necessarily reduce supply chain costs.

SA03.4 Salesforce Incentives & Inventory Management

• Fangruo Chen; Columbia University, Grad. Sch. of Bus., 3022 Broadway, Uris Hall, New York, NY 10027-6902; fchen@research.gsb.columbia.edu

We consider a prevalent salesforce compensation plan that pays commissions on annual sales in excess of a quota. We found that this incentive system induces a nonstationary sales pattern detrimental to the firm's operations. An alternative plan is proposed to induce the salesforce to actually smooth the firm's demand process, generating substantial benefits to the firm.

Simulation Applications I

SA04.1 Learning Simulation on the Web

• Ingolf Stahl; Stockholm School of Economics, Box 6501, Stockholm, S-11383 , Sweden; ingolf.stahl@hhs.se,, http://www.hhs.se/secc/staff/#iis

Since there is a demand in colleges and hgh schools for using simulation in courses and project work, we are putting micro-GPSS, a stream-lined, easy-to-use version of GPSS, on the Web with a GUI for students to use to build up their models by clicking on block symbols.

SA04.2 Web-Based Simulation Modeling using Java

• Vakulathil Abdurahiman; CTA-ITA-IEC, Div. of Computer Sci., Sao Jose dos Campos, SP 12228-900 , Brazil; rahiman@comp.ita.cta.br,, http://www.comp.ita.cta.br/~rahiman
• Celso Hirata; CTA-ITA-IEC, Div. of Computer Sci., Sao Jose dos Campos, SP 12228-900 , Brazil; hirata@comp.ita.cta.br,, http://www.comp.ita.cta.br/~hirata
• Germano Kienbaum; LAC-INPE, Sao Jose dos Campos, SP 12201-970 , Brazil; germano@lac.inpe.br

We discuss the concepts of Web-based simulation modeling. The issues in simulation modeling with adopting this new technique are analyzed. An application example in Java is presented to show how the new simulation environment will help teach simulation.

SA04.3 Maintenance Workforce Optimizaiton in a Cable Factory

• Mehmet Savsar; Kuwait University, College of Eng. & Petroleum, PO Box 5969, Safat, 13060 , Kuwait; mehmet@kuc01.kuniv.edu.kw

We describe a simulation modeling procedure to determine the optimum maintenance workforce level in a cable manufacturing factory with 23 machines. The model is experimented with single-skilled and multi-skilled maintenance crews and the optimum number of repairmen is determined in each case. The cost results are then compared.

SA04.4 UML as Domain Specific Language for the Construction & Maintenance of Simulation-Models in the Area of Circuit-Board Production

• Alpetkin Erkollar; University of Klagenfurt, Dept. of Comp. Sci., Klagenfurt, A-9020 , Austria; alp@ifit.uni-klu.ac.at
• Alexander Felfernig; University of Klagenfurt, Dept. of Comp. Sci., Klagenfurt, 9020 , Austria; alpf@ifit.uni-klu.ac.at

Many domains require the rapid development of highly adaptable simulation-models. In order to accomplish this task, domain specific modeling concepts for building simulation models are employed. We show how modeling concepts of the simulation area can be introduced in UML and be automatically translated into simulation models...

Decision Analysis & Computational Challenges

SA05.1 The Value of Control Revisited

• Ross D. Shachter; Stanford University, EES & OR Dept., Terman Ctr., Stanford, CA 94305-4023; shachter@stanford.edu
• David Heckerman; Microsoft Research, One Microsoft Way, 9S/1024, Redmond, WA 98052; heckerma@microsoft.com

The value of control is the amount a decision-maker should be willing to pay to transform an uncertainty into a decision, that is, to choose its state. Using recent results in causal modeling, we reexamine the value of control and how to compute and apply it.

SA05.2 Efficient Representations for Aggregate Belief & Compact Securities Markets

• David M. Pennock; University of Michigan, UM AI Lab., 1101 Beal Ave., Ann Arbor, MI 48109-2110; dpennock@umich.edu
• Michael P. Wellman; University of Michigan, 427 Riverview Dr., Ann Arbor, MI 48104; wellman@umich.edu

Bayesian networks can exploit conditional independence to compactly represent individual belief and efficiently compute decision-theoretic queries over large domains. We report both positive and negative results bearing on the possibility of exploiting conditional independence to achieve similar efficiency in models of consensus belief and general market equilbrium under uncertainty.

SA05.3 One Practitioner's Perspective on Decision Analysis & its Computational Challenges

• Kazuo J. Ezawa; Mondex International Ltd., Atlantic Tech. Ctr., Ste. 109, 100 Campus Dr., PO Box 972, Florham Park, NJ 07932-0972; kazuo.ezawa.mondex@worldnet.att.net

Using some examples, the reality of computational challenge is reviewed. The more computational power/capability doesn't necessarily mean the better DA. We will discuss trade off issues such as framing, model complexity, seeking insights vs. accuracy of prediction, strategic vs. tactical decisions and ease of use of DA software.

SA05.4 Leveraging Probability & Utility in Computational Systems & Applications

• Eric Horvitz; Microsoft Research, Decision Theory Group, One Microsoft Way, Redmond, WA 98052-6399; horvitz@microsoft.com

I will present research on the use of decision-analytic concepts for making automated decisions about the allocation of processing and memory resources in computer systems. The methods highlight challenges and opportunities in the realm of harnessing representations of probability and utility to enhance the behavior of computer systems and applications.

Very Large-Scale Neighborhood Search

SA06.1 Very Large-Scale Neighborhood Search

• James B. Orlin; MIT, Sloan Sch. of Mgmt., E40-147, Cambridge, MA 02139; jorlin@mit.edu
• Ravindra K. Ahuja; University of Florida, 303 Well Hall, Dept. of ISE, Gainesville, FL 32611; ahuja@ise.ufl.edu
• Ozlem Ergun; MIT, OR Ctr., 77 Mass Ave., Bldg. E40-130, Cambridge, MA 02139; ozie@mit.edu
• Abraham P. Punnen; University of New Brunswick, Dept. of Math. Stats. & CS, PO Box 5050, Saint John, NB, E2L 4L5 , Canada; punnen@unbsj.ca

In neighborhood search techniques, one searches the 'neighborhood' of a given solution x' and replaces x' by a neighbor that has improved cost. VLSN search refers to neighborhood search techniques in which the number of neighbors of x' may be exponentially large. We briefly survey VLSN search.

SA06.2 Searching Very Large-Scale Neighborhoods using Improvement Graphs

• Ravindra K. Ahuja; University of Florida, 303 Well Hall, Dept. of ISE, Gainesville, FL 32611; ahuja@ise.ufl.edu
• James B. Orlin; MIT, Sloan Sch. of Mgmt., E40-147, Cambridge, MA 02139; jorlin@mit.edu

We review 2 applications of very large-scale neighborhood search in which the key idea is to search neighborhoods using improvement graphs. We illustrate the ideas on partitioning problems, especially to the capacitated minimum spanning tree problem, and on airline fleet assignment.

SA06.3 Large-Scale Neighborhood Search Applied to a Nonbifurcated Network Loading Problem with Multiple Facilities

• Bernard Gendron; Universite de Montreal, DIRO & CRT, CP 7128, succ. Centre-ville, Montreal, Quebec, H3C 3J7 , Canada; bernard@crt.umontreal.ca
• Jean-Yves Potvin; Universite de Montreal, DIRO/CRT, CP 6128, Succ, Centre-Ville, Montreal, Quebec, H3C 3J7 , Canada; potvin@diro.umontreal.ca
• Patrick Soriano; University of Montreal, HEC, CP 6128, succ. Centre-ville, Montreal, Quebec, H3C 3J7 , Canada; partick@crt.umontreal.ca

We study a network design problem that arises in the telecommunications industry. We describe large-scale neighborhood search heuristics, which perform k-opt type exchanges in the space of paths. Numerical results are presented on different types of networks with up to 200 vertices.

SA06.4 Domination Analysis of Some Well-Known Construction Heuristics for the Traveling Salesman Problem

• Abraham P. Punnen; University of New Brunswick, Dept. of Math. Stats. & CS, PO Box 5050, Saint John, NB, E2L 4L5 , Canada; punnen@unbsj.ca

We review domination results on several well known construction heuristics for the TSP including the Christofides heuristic, patching algorithms, nearest neighbor and other greedy type algorithms, node insertion algorithms, etc. Some new results in this area will also be presented.

Panel: Research Directions in Manufacturing Logistics

SA07.1 Panel: Research Directions in Manufacturing Logistics

• Robin O. Roundy; Cornell University, Dept. of OR/IE, 203 Rhodes Hall, Ithaca, NY 14853; robin@orie.cornell.edu
• David Simchi-Levi; Northwestern University, Dept. of IE/MS, 2145 Sheridan Rd., Evanston, IL 60208; levi@iems.nwu.edu
• Guillermo Gallego; Columbia University, IEOR Dept., Mudd Bldg., 500 West 120th St., Rm. 331, New York, NY 10027; ggallego@ieor.columbia.edu
• Izak Duenyas; University of Michigan, Dept. of IOE, 1205 Beal Ave., Ann Arbor, MI 48109; duenyas@umich.edu
• S. David Wu; Lehigh University, Dept. of IMSE, 200 West Packer Ave., Bethlehem, PA 18015; david.wu@lehigh.edu

Research in manufacturing logistics expands the traditional scope of production and operations management to include issues such as supply chain coordination, information technology, and the integration of manufacturing and other business functions. This panel will discuss recent and future trends for research in this area.

Modeling the Supply Chain at Lucent Technologies: Practitioners View of the Supply Chain Problem

SA08.1 FSPAT: A Tool for Targeting & Reducing Significant Cost Elements in the Supply Chain

• Robert R. Svenson; Bell Labs., Lucent Technologies, PO Box 900, Princeton, NJ 08542-0900; svenson@lucent.com
• Bruce H. Yuan; Bell Labs., Lucent Technologies, PO Box 900, Princeton, NJ 08542-0900; byuan@lucent.com

Many of the most important decisions affecting a manufacturing firm engaged in international business require a complete understanding of the supply chain and its significant cost elements. To support Lucent Technologies' planning activities, a financial accounting tool was developed to support scenario-based decision analysis. The tool will be described with example applications.

SA08.2 Global Virtual Factory & Supply Chain Modeling in a Build-to-Order Environment

• Turgut Aykin; Bell Laboratories, MSE Dept., Eng. Research Ctr., PO Box 900, Princeton, NJ 08542; taykin@lucent.com

We describe a supply chain planning model for the management of material flows in a build-to-order environment where performance metrics are specified not in terms of item level fill rates, but in order level fill rates, and order delivery lead time and reliability. We also present a framework for a global virtual factory involving multiple factories, distribution centers, markets and suppliers.

SA08.3 The Integration of Competitive Analysis, Market Information & Reverse Engineering: Methods to Determine Competitor Supply Chain Strategies

• Lee L'Esperance; Lucent Technologies, PO Box 900, Princeton, NJ 08081; llesperance@lucent.com

The supply chain is becoming an increasingly important element of competitive strategy today. Contract manufacturing, third-party logistics, web-based management are among some of the methods used to maximize manufacturing assets while minimizing product delivery intervals. Methods will be described that can lead to insights in how competitor's supply chains link to their overall strategy and financial performance.

SA08.4 Collaborative Planning at Lucent Technologies

• Mary J. Meixell; Bell Labs., Lucent Technologies, PO Box 900, Princeton, NJ 08542-0900; marymeixell@lucent.com
• Daniel Bilodeau; Bell Labs., Lucent Technologies, PO Box 900, Princeton, NJ 08542-0900; bilodeau@lucent.com

Collaborating along the supply chain leads to improvement in both operational and ultimately financial performance of the enterprise. We review how the benefits associated with collaborative planning are estimated.

AI Applications

SA09.1 Intelligent Data Mining Using Object-Oriented Paradigm

• Davinder K. Malhotra; Philadelphia College, Sch. House Ln. & Henry Ave., Philadelphia, PA 19144; malhotrad@philacol.edu
• Rashmi Malhotra; St. Joseph's University, Dept. of MIS, 5600 City Ave., Philadelphia, PA 19131;

Although many different intelligent techniques, i.e., neural network methods, GAs and neuro-fuzzy models, etc., have been applied for data mining and data warehousing, knowledge-based systems have not been very extensively applied to the DSS data analysis. We illustrate the use of artificially intelligent knowledge-based methods in retrieving information from the morningstar's mutual fund database.

SA09.2 A Geometric Model for Computational Music Analysis

• Elaine Chew; MIT, 158 Webster Ave., Cambridge, MA 02141; eniale@mit.edu,, http://web.mit.edu/eniale

Enjoyment of the arts is a defining feature of an enhanced quality of life. The quest for deeper understanding and appreciation naturally follows. Computational tonal analysis is a difficult problem which has been tackled with limited success in music cognition. My new approach utilizes a geometric model which incorporates functional pitch relations.

SA09.3 A Generalized Neural Network Approach to Chemical Oxidation Process Modeling

• Sarah S. Lam; Binghamton University, Dept. of Systems Sci. & IE, Binghamton, NY 13902; sslst7@pitt.edu
• Alice E. Smith; Auburn University, ISE Dept., 207 Dunston Hall, Auburn, AL 36832; aesmith@eng.auburn.edu

An approach that considers multiple feed gas types and reactor types for a chemical oxidation process is presented. In contrast to a more traditional approach, quadratic response surface modeling, our approach requires the construction of one generalized model, which can be used for optimizing the process...

SA09.4 Sensitivity Tests on a Set of Learning Classifier Systems Trained to Play the Iterated Prisoner's Dilemma Problem

• Kenneth S. Sims; Cumberland College, 7656 College Station Dr., Williamsburg, KY 40769-1387; kenpc@cc.cumber.edu
• Ram Pakath; University of Kentucky, DSIS, Sch. of Mgmt., 425D Gatton B&E Bldg., Lexington, KY 40506-0034; pakath@ukc.uky.edu

We extend prior explorations that use LCS-based agents to generate superior IPD game-playing strategies against opponents playing human-proposed strategies. We provide a 'sensitivity analysis' of the LCS by observing its reactions to alterations in some of its parameters and to some opponent parameters.

Semidefinite Programming

SA10.1 Interior-Point Methods for Semidefinite Programming

• Michael J. Todd; Cornell University, School of OR/IE, 229 Rhodes Hall, Ithaca, NY 14853-3801; miketodd@cs.cornell.edu,, http://www.orie.cornell.edu/~miketodd/todd.html
• Kim C. Toh; National University of Singapore, Dept. of Math., 10 Kent Ridge Crescent, Singapore, 0511 , Singapore; mattohkc@math.nus.sg
• Reha H. Tutuncu; Carnegie Mellon University, Dept. of Math. Sci., 6113 Wean Hall, Pittsburgh, PA 15213; reha@andrew.cmu.edu

We discuss some issues related to interior-point methods for semidefinite programming, including the work involved at each iteration and the ability to exploit sparsity and special structure. We give computational results for the code SDPT3 of Toh, Todd & Tutuncu on problems from SDPLIB-1.1.

SA10.2 A Polynomial-Time Algorithm for Semidefinite Optimization Based on a Scaled Gauss-Newton Direction

• Cornelis Roos; Delft University of Technology, Fac. of Info. Tech. & Systems, Mekelweg 4, PO Box 5031, Delft, 2628 CD , The Netherlands; c.roos@its.tudelft.nl
• Etienne de Klerk; Delft University of Technology, Faculty TWI/SSOR, Mekelweg 4, Delft, 2628CD , The Netherlands; e.deklerk@its.tudelft.nl
• Jiming Peng; Delft University of Technology, Fac. of Info. Tech. & Systems, PO Box 5031, Delft, 2600 GA , The Netherlands; j.peng@its.tudelft.nl
• Tamas Terlaky; Delft University of Technology, Fac. of Info. Tech. & Systems, Mekelweg 4, Delft, 2628 CD , The Netherlands; t.terlaky@its.tudelft.nl

Inspired by a paper of Kruk et al., we propose a new primal-dual search direction for solving SDO problems. The search direction is scale invariant, coincides with the primal-dual direction in the linear case and yields a polynomial-time method when subsequent targets on the central path are chosen appropriately.

SA10.3 Using Semidefinite Programming to Obtain a New Quadratic Programming Relaxation of the Quadratic Assignment Problem

• Kurt M. Anstreicher; University of Iowa, Dept. of MS, Iowa City, IA 52242; kurt-anstreicher@uiowa.edu

It has been shown that a simple eigenvalue bound for the QAP has a semidefinite programming representation. We use this SDP representation to construct a new convex qp relaxation of QAP. The lower bound obtained from this QP is stronger than the well-known projected eigenvalue bound for QAP.

Discrete & Nonconvex Optimization Applications

SA11.1 Reformulation Technique for Nonlinear Generalized Disjuctive Programs

• Ignacio E. Grossmann; Carnegie Mellon University, Dept. of Chem. Engineering, Pittsburgh, PA 15213; grossmann@cmu.edu
• Sangbum Lee; Carnegie Mellon University, Dept. of Chem. Eng., Pittsburg, PA 15213; sangbum@andrew.cmu.edu

We address the solution of generalized disjunctive programs, which involve Boolean and continuous variables, disjunctions containing convex nonlinear inequalities and logic propositions. A reformulation method is proposed based on determining the convex hull of each disjunction, that allows the application of algebraic or logic based MINLP methods.

SA11.2 Reactive Scheduling under Uncertainty Considerations for Multiproduct Batch Plants

• Marianthi G. Ierapetritou; Rutgers University, Dept. of Chem. & Biochem. Eng., 98 Brett Rd., Piscataway, NJ 08854-8058; marianth@sol.rutgers.edu

We address the problem of considering uncertainty into the detail short-term scheduling problem. Uncertainty can be viewed as unexpected deviations in unit availabilities (machine break-down) and/or fluctuations in product demands and raw material availabilities. The approach is based on Ierapetritou & Floudas', 1998 model and determines a new optimal schedule...

SA11.3 Midterm Supply Chain Planning under Demand Uncertainty

• Anshuman Gupta; Pennsylvania State University, Dept. of Chem. Engineering, 112 Fenske Lab., University Park, PA 16802; anshuman@santorini.che.psu.edu
• Costas D. Maranas; Pennsylvania State University, 112A Fenske Lab., University Park, PA 16802; costas@psu.edu

A 2-stage, stochastic programming approach is proposed for incorporating demand uncertainty in midterm planning problems. Based on duality theory, a closed form solution for the recourse problem is obtained followed by expectation evaluation by analytical integration over the multivariate probability space. The resulting convex MINLP is solved to optimality with the outer approximation algorithm.

Decision Making

SA12.1 'Short-Termism' in New Product Decisions

• William T. Ross, Jr.; Temple University, Sch. of Bus. & Mgmt., 1810 North 13th St., Philadelphia, PA 19122;
• Marjorie Adams; University of Virginia, Darden School, PO Box 6550, Charlottesville, VA 22906;

There is evidence that managers will overvalue short-term projects at the expense of longer-term options. We conduct 3 studies. First, we examine, in a new product situation, whether managers do indeed favor short-term projects. Secondly, we study whether managers who adopt a portfolio management approach to the management of projects are less likely to demonstrate this bias...

SA12.2 Predicting an Uncertain Future to Make a Decision Now: Heuristics & Biases in Assessing Extreme-Value Distributions

• Wes Hutchison; University of Pennsylvania, The Wharton School, Pittsburgh, PA 19104;
• Robert J. Meyer; University of Pennsylvania, The Wharton School, Mktg. Dept., Philadelphia, PA 19104-6371;

Central to many dynamic decisions faced by consumers and managers is an intuitive assessment of the expected value of an uncertain future choice. That choice will be the best option available at the time, and it can be represented as the maximum (or sometimes the minimum) of a finite number of draws from some distribution...

SA12.3 Framing Effects of Market Structure in Judgments of Health Risk & Message Processing

• Lauren Block; NYU, 44 West 4th St., New York, NY 10012;
• Geeta Menon; NYU, Stern Sch. of Business, 44 West 4th St., Rm. 8-94, New York, NY 10012;

We examine the effects of risk-factor framing and the number of risk factors enumerated in an ad on people's perceptions of risk, attitudes and behavioral intentions toward hepatitis C testing. Results from two studies demonstrate the interactive effects of risk type and number of risk factors on health persuasion.

SA12.4 Coping with Bad Choices: Attribute Value & Emotional Tradeoff Difficulty

• Mary Frances Luce; University of Pennsylvania, The Wharton Sch., Philadelphia, PA 19104;
• James R. Bettman; Duke University, Fuqua School of Bus., Box 90120, Durham, NC 27708-0120;
• John W. Payne; Duke University, Fuqua School of Bus., Durham, NC 27706;

We investigate the hypothesis that decision-makers will be motivated to cope with the increasing threat presented by low attribute values by maximizing the relevant attribute in choice. We report 2 studies that use choice patterns and emotion ratings to isolate this coping behavior, controlling for several additional sources of influence on choice (e.g., influence associated with attribute importance weights).

Distribution Channels

SA13.1 withdrawn - author request of 10/11

• Steffen Jorgensen; Odense University, Dept. of Mgmt., Campusvej 55, Odense, DK-5230 , Denmark; stj@sam.sdu.dk
• Georges Zaccour; Ecole des HEC, Dept. of Mktg., 3000 Chemin Cote Ste-Catherine, Montreal, Quebec, H3T 2A7 , Canada; georges.zaccour@hec.ca

SA13.2 Demand Externalities & Interactions on the Quality of Services in Franchise Chains

• Janghyuk Lee; IHEL of ESSEC, Avenue Bernard Hirsch, BP 105 95021, Cergy Pontoise, , France;

We focus on the consequences of interactions between 2 franchised outlets on service quality as demand externalities exist in the chain. Using a formal model, we present the form of optimal reaction function and analyze different equilibria of service quality in terms of information availability. Empirical findings from 2 major companies in France will be also presented.

Technology & Networks

SA14.1 The Quality of Collaboration: Inducements & Opportunities in the Formation of Interfirm Linkages

• Gautam Ahuja; University of Texas, Management Dept., CBA 4.202, Austin, TX 78712; ahujaebos@texas.edu

I argue that centrality in an inter-firm network is explained by simultaneously examining both linkage-inducement and linkage-opportunity factors. Firms possessing 3 inducement and opportunity factors accumulated technical, commercial and social capital and become central more easily. However, firms lacking these accumulated resources can also become central by creating radical tehnological breakthroughs.

SA14.2 Coming From Good Stock: Career Histories & New Venture Formation

• M. Diane Burton; Harvard Business School, Soldiers Field Rd., Boston, MA 02163; dburton@hbs.edu
• Jesper B. Sorensen; University of Chicago, Grad. Sch. of Bus., 1101 East 58th St., Chicago, IL 60637; jesper.sorensen@gsb.uchicago.edu
• Christine Beckman; Stanford University, Grad. Sch. of Bus., Stanford, CA 94305; cbeckman@gsb.stanford.edu

We argue that entrepreneurs secure informational and reputational benefits from having worked for employers who occupy prominent positions in entrepreneurial networks. Employer centrality increases employee exposure opportunities and resources. Employer prominence reflects positively on employees, helping to reduce perceived uncertainty of proposed ventures. We find that Silicon Valley founders pursue risky, innovative ventures at a higher rate...

SA14.3 Strategic Participation in Cooperative Technical Organizations: Informal Networks & Alliance Formation

• Lori Rosenkopf; University of Pennsylvania, The Wharton School, 2000 SH-DH, Philadelphia, PA 19104-6302; rosenkopf@wharton.upenn.edu
• Anca Turcanu; University of Pennsylvania, The Wharton School, 2000 SH-DH, Philadelphia, PA 19104-6370; turcanu@management.wharton.upenn.edu
• Varghese George; Rutgers University, Fac. of Mgmt. Org. Mgmt. Dept., 111 Washington St., Newark, NJ 07102; vgeorge@andromeda.rutgers.edu

We derive an informal inter-firm network firm participation in technical committees in the cellular industry. We show that firm centrality in this network increases subsequent alliance formation rates and that greater joint participation by a pair of firms in this network increases its likelihood of forming subsequent alliances.

SA14.4 The Geography of Opportunity

• Toby Stuart; University of Chicago, Grad. Sch. of Bus., 1101 East 58th St., Chicago, IL 60637; toby.stuart@gsb.uchicago.edu
• Olav Sorensen; University of Chicago, Grad. Sch. of Bus., 1101 East 58th St., Chicago, IL 60637; olav.sorensen@gsb.uchicago.edu

We develop a socialogical explanation for industry clustering. We present a network and ecological perspective on the determinants of spatial heterogeneity in founding rates and organizational performance in the population of US biotechnology firms. Our analyses investigate the effects of geographic proximity to established biotechnology firms, biotechnology experts and venture capitalists on the location-specific founding rates and performance of startups.

Knowledge Management in the Contemporary Firm

SA15.1 The Emergence of Leadership in Knowledge Management: A Look at Forms & Functions

• Patricia D. Witherspoon; University of Texas, Dept. of Speech Comm., Austin, TX;

Knowledge management in contemporary organizations requires new approaches to the exercise of leadership within these organizations. We look at the characteristics and expectations of those engaged in the process of knowledge management as a process and suggest new forms of leadership that emerge in this process.

SA15.2 Profound Knowledge & Software Systems Engineering: Is there a Link?

• Bobby G. Moore; SAIC, 4242 Woodcock Dr., Ste. 150, San Antonio, TX 78228; bmoore@texas.net

Deming's 'profound knowledge' and Checkland's soft system methodology are described and analyzed to determine if a linkage exists between them and if so, what the nature of the linkage is.

SA15.3 Using Technology Transfer's Empirical Research to Inform the Area of Knowledge Management

• Kathy L. Wohlert; University of Texas, Div. of Mgmt. & Mktg., 6900 North Loop 1604 West, San Antonio, TX 78249; kwohlert@utsa.edu

Although there is a burgeoning interest in and implementation of knowledge management programs in many of today's firms, limited empirical research exists in knowledge management. Technology transfer is identified as a subset of knowledge management and the research from the field of technology transfer is applied accordingly.

SA15.4 Knowledge Management & Adding Value to an Organization

• Glenn Dietrich; University of Texas, Div. of Acctng. & Info Sys., 6900 North Loop 1604 West, San Antonio, TX 78249; gdietrich@utsa.edu

Knowledge management has become an important part of business. We explore ways that knowledge management can add value to an organization. Particular emphasis is placed on creating conditions within the firm that foster knowledge management. Using knowledge management as a technique for developing strategic plans in a rapidly changing technology environment is also discussed.

Portfolio Analysis

SA16.1 Problems of Optimization of an Investment Portfolio

• Efim M. Bronshtein; Ufa State Aviation Technical University, K.Marx Str. 12, Ufa, Bashkortostan, 450000 , Russia; brem@soros.basheda.ru
• Semjon I. Spivak; Bashkir State University, Frunze Str. 32, Ufa, Bashkortostan, 450074 , Russia; spivak@bsu.bashedu.ru

Formulations of problems of formation of optimal investment portfolio are shown at various definitions of risk of the projects. The interest rates are assumed to be stochastic. Different approaches are applied such as Markov claims, time-depending probabilities of refusals of investments and indexes of a condition.

SA16.2 Index Fund with Mean Square Tracking

• Yoshio Tabata; Osaka University, Grad. School of Econ., Toyonaka, Osaka, 560 , Japan; tabata@econ.osaka-u.ac.jp

Some properties of index fund designing problems are shown in discrete time and continuous time. The main property is that if the benchmark is (not) on the efficient frontier, then the index fund with minimum tracking error is (not) on the efficient frontier. More interesting results will be provided.

SA16.3 A Dynamic Asset Allocation Model with Downside Risk Control

• Yonggan Zhao; University of British Columbia, Fac. of Commerce, #419 2053 Main Mall, Vancouver, BC, V6T 1Z2 , Canada; zhao@phdlab.commerce.ubc.ca
• William T. Ziemba; University of British Columbia, Fac. of Commerce, 2053 Main Mall, Vancouver, BC, V6T 1Z2 , Canada; ziemba@interchange.ubc.ca

Assuming lognormality for prices, the strategy that gauges the dynamic portfolio weight by the risk neutral excess rate of return is determined by a stochastic differential equation. A constrained optimization model is established given investors' risk preference and asset price model. Under the risk measure, value at risk, the downside control method is superior to both buy and hold and fixed mix strategies.

SA16.4 Mean-Variance Portfolio Selection with Volatility Degeneracy Case

• Xun Li; Chinese University of Hong Kong, Dept. of SEEM, Shatin NT Hong Kong, , China; xli@s.cuhk.edu.hk,, http://www.se.cuhk.edu.hk/~xli/
• Xunyu Zhou; Chinese University of Hong Kong, Dept. of SEEM, Shatin NT Hong Kong, , China; xyzhou@se.cuhk.edu.hk,, http://www.se.cuhk.edu.hk/~xyzhou/

The mean-variance portfolio selection model with volatility degeneracy is studied in a financial market containing 1 bond and m stocks whose prices are associated with a d dimensional Brownian motion process. The efficient frontier, along with the analytical efficient policies, is obtained via solving an auxiliary stochastic linear-quadratic problem.

The Role of Information in Various Forms of Scheduling

SA17.1 Interaction between Supply Chain Planning & Control & Shopfloor Scheduling in Parallel Flow Shops

• A. G. de Kok; Eindhoven University of Technology, Paviljoen F4, PO Box 513, Eindhoven, 5600 MB , The Netherlands; a.g.d.kok@tm.tue.nl

We present the planning and control of a supply chain where a particular link can be modeled as a parallel flow shop. At the supply chain level, we use a rolling schedule LP model to plan capacity and material availability. At the shop level, we apply a local search algorithm to determine the detailed production plan...

SA17.2 The Use of Foreknowledge in Order Acceptance

• Jacob Wijngaard; University of Groningen, Assen, 9406 CN , Netherlands; winjngaard@bdk.rug.nl

Order acceptance is an important decision level in production control. One approach is to direct order acceptance by actual detailed production schedules. The strengths and weaknesses of this approach will be discussed and the approach will be compared with a more aggregate approach. The influence of tacit elements in customer preferences will be considered.

SA17.3 Queueing Models for Capacitated Multi-Echelon, Multi-Indenture Repairable Item Systems

• W. H. M. Zijm; Eindhoven University of Technology, PO Box 513, Paviljoen F18, Eindhoven, 5600 MB , The Netherlands; w.h.m.zijm@tm.tue.nl

We discuss the performance analysis of multi-product, multi-echelon, multi-indenture inventory systems for repairable items. At the bases (downstream) and at a central depot, both repair facilities are available to upgrade failed parts for reuse. All repair facilities are subject to capacity constraints. We develop integrated production/inventory models, based on queueing networks and discuss analytical and numerical results.

SA17.4 Using Workload & Routing Information to Reduce Lead Time Uncertainty in A Job Shop Production System

• Jan C. Fransoo; Eindhoven University of Technology, PO Box 513, Pav F12, Eindhoven, NL-5600 MB, The Netherlands; j.c.fransoo@tue.nl,, http://www.tm.tue.nl/vakgr/lbs/fransoo.htm
• H. P. G. van Ooijen; Eindhoven University of Technology, Eindhoven, 13970 , The Netherlands;

We present a statistical model that relates order throughput times to the workload at the work centers on the routing of a job under different shop floor dispatching rules. Based on the statistical model, a prediction model is derived that assigns cost optimal external due dates to jobs per routing class.

Combinatorial Optimization Solutions for Applications in Industry

SA18.1 Integer Knapsack-Cover Polyhedra & Applications

• David R. Mazur; Johns Hopkins University, Dept. of Math. Sci., 220 Maryland Hall, Baltimore, MD 21218; mazur@mts.jhu.edu
• Leslie A. Hall; Johns Hopkins University, Dept. of Math. Sciences, Baltimore, MD 21218; lah@jhu.edu

We present structural results on polyhedra related to the integer knapsack-cover problem. We then discuss some computational implications of these results and report on their effectiveness in a cutting-plane algorithm for solving a particular telecommunications planning problem.

SA18.2 Optimizing Paint Blocking in an Automobile Assembly Line: An Application of Specialized TSPs

• Joel Sokol; MIT Operations Research Center, 305 Memorial Drive, Cambridge, MA 02139; jsokol@mit.edu
• Thomas L. Magnanti; MIT, Sloan Sch. of Mgmt., Dept. of Elect. Eng. & CS, Cambridge, MA 02139-4307; magnanti@mit.edu

In automobile assembly plants, painting similarly-colored cars sequentially saves millions of dollars annually. We model this problem as a time-window TSP. We compare various formulations containing up to 14,000,000 variables for real instances, and exploit the problem's structure to obtain good, fast solutions that allow real-time adjustments.

SA18.3 Inferring Network Topology from Packet Delay Data

• David F. Shallcross; Bellcore, 445 South St., 1J-306B, Morristown, NJ 07960; davids@bellcore.com

We consider a system of network monitoring by exchanging packets over a period of time between selected points. Using packet delay data, we want to determine the topology of the network, and the routing of the packets. We describe the methods we have tried and their performance on test problems.

SA18.4 Heuristics for Routing & Slotting on Ring Networks

• Steven T. Cosares; Dowling College, Dept. of Math, Oakdale, NY 11769; cosares@dowling.edu
• Tami Carpenter; Bellcore Room 1A-320B, 445 South St., Morristown, NJ 07960; tcar@bellcore.com
• Iraj Saniee; Bell-Labs, Lucent Technologies, 600 Mountain Ave., Murray Hill, NJ 07974; iis@research.bell-labs.com

We describe the performance of heuristics for a routing and slotting problem that arises in SONET ring sizing. We consider both sequential approaches, wherein the routing of demands precedes their assignment to slots, and more integrated approaches that consider the route and slot assignments simultaneously.

Hybrid Algorithms

SA19.1 Exact Solutions for Real World Crew Scheduling Problems

• Arnaldo V. Moura; ; arnaldo@dcc.unicamp.br
• Cid C. de Souza; UNICAMP, Instituto de Computacao, CP 6176, Campinas, Sao Paulo, 13083-970 , Brazil; cid@dcc.unicamp.br
• Tallys Hoover Yunes; ; tallys@acm.org

We present a hybrid methodology for optimally solving large-scale crew scheduling problems. These are real world instances reaching more than 1.5 billion entries. Our approach integrates constraint programming and integer programming, exploiting their particular strengths for solving specific parts of the problem. Implementation details and computational experiments are discussed.

SA19.2 Hybrid Algorithms for the 'Oregon Trail' Knapsack Problem

• Jennifer Burg; Wake Forest University, PO Box 7388; burg@mthcsc.wfu.edu
• Sheau-Dong Lang; University of Central Florida, School of Computer Sci., Orlando, FL 32816; lang@cs.ucf.edu

The Oregon Trail knapsack problem imposes a cost as well as a weight limit, constraining the values of items by a variety of value functions, possibly depending on the presence or absence of other items. The solution combines dynamic programming, constraint propagation techniques and classic B&B approaches.

SA19.3 Forward Search Hybridization Forms

• Hani El Sakkout; Imperial College, IC-Parc, William Penney Lab., London, SW7 2AZ , England, UK; hhe@icparc.ic.ac.uk

Forward search is an enhancement of backtracking implicit in many hyprid search algorithms, whereby an independent search procedure is applied to a sub-problem to reduce the subsequent backtrack search. The forward search mechanisms in several hybrid algorithms are considered, and two principal forms are identified, namely hierarchical commitment and probe backtracking.

SA19.4 A Hybrid CP-OR Approach to Solve the Traveling Salesman Problem with Time Windows

• Filippo Focacci; Universite di Ferrara, Dipt. di Ingeneria, Via Saragat 1, Ferrara, 44100 , Italy; ffocacci@deis.unibo.it
• Andrea Lodi; University of Bologna, DEIS, Viale Risorgimento 2, Bologna, 10136 , Italy;
• M. Milano; University of Bologna, DEIS, Viale Risorgimento 2, Bologna, 10136 , Italy; mmilano@deis.unibo.it

We developed constraints for the TSPTW, integrating OR bounding techniques and dynamic programming, that filter domains to remove inconsistent values and non-optimal ones. Embedding these constraints in constraint programming yields a model that is flexible, and efficient. The behavior of our implementation, using ILOG Solver, is compared with related approaches.

Applications for Urban Transit Systems

SA20.1 A Method to Adjust Passenger Counts on a Transit Line to Achieve Consistency

• Shinya Kikuchi; University of Delaware, Dept. of Civil & Environ. Eng., Newark, DE 19716-3122; kikuchi@ce.udel.edu

Boarding and alighting passenger counts on a transit line is subject to large amount of error. Hence, it is necessary to adjust the original counts at each station so that the conservation of flow (total boarding = totaling alighting) must be satisfied. Traditionally, this task has been done in an ad hoc manner...

SA20.2 Real-Time Scheduling of Demand Responsive Transit Systems

• Maged Dessouky; University of Southern California, Dept. of ISE, Los Angeles, CA 90089-0193; maged@rcf.usc.edu
• Randolph W. Hall; University of Southern California, Dept. of ISE, Los Angeles, CA 90089-0193;
• Guntram Noeth; University of Southern California, Dept. of ISE, Los Angeles, CA 90089-0193;
• Stefan Adam; University of Southern California, Dept. of ISE, Los Angeles, CA 90089-0193;

With the passage of the ADA Act, which requires that transit agencies provide para-transit or on demand service for the disabled, there has been renewed interest in demand responsive transit. In this talk, we present a heuristic for real-time scheduling of such systems. We evaluate the heuristic using data provided by para-transit service providers in Los Angeles County.

SA20.3 Operations Research in Transit

• Vukan R. Vuchic; University of Pennsylvania, Dept. of Sys. Eng., 220 South 33rd St., Rm. 293, Philadelphia, PA 19104-6315; vuchich@seas.upenn.edu

Although urban transit operations are conducive to OR applications, achievements are mixed: some useful work, but many studies without real-world applications. Insufficient contacts between theoreticians and practitioners and unrealistic models are common problems. Critical review will be followed by suggestions for increased OR applications in transit.

SA20.4 Techniques for Scheduling Integrated Transit Service

• Mark Hickman; Texas A&M University, Dept. of Civil Eng., College Station, TX 77843; mhickman@tamu.edu
• Kelly Blume; Texas A&M University, Dept. of Civil Eng., College Station, TX 77843-3136;

Demand-responsive transit service may be integrated with fixed route service to carry passengers on longer trips. We present a modified trip itinerary algorithm to schedule these passenger trips. The use of the algorithm and a vehicle scheduling technique are investigated using a case study in Houston.

Dynamic Traffic Assignment I

SA21.1 Computational Experience with Large-Scale Dynamic Traffic Assignment Models

• S. Travis Waller; Northwestern University, Dept. of Civ. Eng., 2145 Sheridan Rd., Tech. Inst., Evanston, IL 60208; travis@trans.civil.nwu.edu
• Athanasis Ziliaskopoulos; Northwestern University, 2145 Sheridan Rd., Evanston, IL 60208; a-z@nwu.edu
• Yue Irene Li; Northwestern University, Dept. of ISE, 2145 Sheridan Rd., Tech. Inst., Evanston, IL 60208; liyue@nwu.edu

A framework for performing large-scale simulation-based DTA on realistically sized networks is presented and discussed. Implementation details for both the simulator and DTA framework are analyzed. Computational results are presented from applying this framework to a 16,000-link network of Columbus, OH.

SA21.2 System Consistency Control for Real-time Dynamic Traffic Assignment Operations

• Ying Kang; University of Texas, Dept. of Civil Eng., ECJ 6.2, Austin, TX 78712-1076; yingkang@mail.utexas.edu
• Hani S. Mahmassani; University of Texas, Dept. of Civil Eng., ECJ 6.2, Austin, TX 78712-1076;

The developed diagnostic architecture of system consistency control for real time DTA is presented along with a demand prediction feedback adjustment algorithm and a simulator adjustment algorithm. Possible research directions and potential applications of this methodology are also discussed.

SA21.3 A Threshold-Constrained System Optimal Dynamic Assignment for Route Guidance Applications

• Hossein Tavana; University of Texas, Dept. of Civ. Eng., ECJ 6.2, Austin, TX 78712; hossein@mail.utexas.edu
• Hani S. Mahmassani; University of Texas, Dept. of Civil Eng., ECJ 6.2, Austin, TX 78712-1076;

A formulation and algorithm are presented for a system-optimal assignment of vehicles to paths in a congested network. The results satisfy the requirements of a boundedly-rational user equilibrium by keeping the travel times along the used paths within users' indifference thresholds.

Negotiation Fundamentals

SA22.1 Negotiations Theory for IT Contracting

• Rachel Croson; University of Pennsylvania, Wharton Sch. of Bus., Dept. of OPIM, 1322 SH-DH, Philadelphia, PA 19104-6366; crosonr@wharton.upenn.edu

No abstract supplied.

SA22.2 Lessons Learned from Facilitating IT Outsourcing Negotiations

• Elizabeth Gray; ;

No abstract supplied.

Decomposition Methods & Applications

SA23.1 Extracting Partial Nonlinear Information from Modeling Languages

• Emmanuel Fragniere; University of Lausanne, HEC, BFSH1, Dorigny Lausanne, CH-1015 , Switzerland; emmanuel.fragniere@hec.unil.ch
• Christian Condevaux-Lanloy; Universite de Geneve, HEC, 102 Boulevard Carl-Vogt, Geneve, , Switzerland; condevea@uni2a.unige.ch

We will present a tool to access and exploit nonlinear parts of a structured model that is generated from an algebraic modeling language. This feature allows to decompose the convex nonlinear programming problem when too large to be solved frontally.

SA23.2 An Interior-Point Cutting-Plane Algorithm for a Lagrangean Decomposition of the Production-Distribution System Design Problem

• Samir Elhedhli; McGill University, Fac. of Mgmt., 1001 Sherbrooke West, Montreal, Quebec, H3A 1G5 , Canada; elhedhls@management.mcgill.ca
• Jean-Louis Goffin; McGill University, Faculty of Mgmt., 1001 Sherbrooke West, Montreal, Quebec, H3A 1G5 , Canada; goffin@management.mcgill.ca

We consider an extension to the analytic centre cutting plane method (ACCPM). We suggest the use of primal, dual and primal-dual interior plant algorithms in the calculation of analytic centers. We explore the motivating theory and discuss the algorithmic issues. We then apply this to a Lagrangean decomposition of the production-distribution system design problem...

SA23.3 Uses of Decomposition Methods in Solving Electric Power Generation Problems

• John R. Birge; University of Michigan, Dept. of IOE, 1205 Beal Ave., Ann Arbor, MI 48109-2117; jrbirge@engin.umich.edu

Deregulated electric power markets present new challenges to the optimal generation of power. We will consider models for the optimal coordination of plants with limited capacity and market power and their use in a decomposition of the market equilibrium problem.

SA23.4 Parallel Primal-Dual Decomposition

• J. David Fuller; University of Waterloo, Dept. of MS, Waterloo, Ontario, N2L 3G1 , Canada; dfuller@engmail.uwaterloo.ca
• H. Jin Park; University of Waterloo, Dept. of MS, Waterloo, Ontario, N2L 3G1 , Canada; hjpark@engmail.uwaterloo.ca

We describe the principles of the parallel primal-dual decomposition method, in which the traditional master-subproblem relationship is replaced by a scheme in which every subproblem behaves like either a Dantzig-Wolfe or Benders master problem with respect to the other subproblems. Some computational results are presented.

Management Issues in Telemedicine

SA24.1 Assessing Technological Barriers to Telemedicine: Technology Management Implications

• David L. Paul; Georgetown University, McDonough School of Bus., Washington, DC 66045; dlp@gusun.georgetown.edu
• Keri E. Pearlson; University of Texas, Dept. of MSIS, Grad. Sch. of Bus., Austin, TX 78712; kpearlson@mail.utexas.edu
• Reuben R. McDaniel; University of Texas, Dept. of MS/IS, Grad. Sch. of Bus., Austin, TX 78712; rmcdaniel@mail.utexas.edu

Drawing on data collected from 3 operational telemedicine projects involving different clinical telemedicine applications, we examine how technological barriers to telemedicine are impacting its utilization rates.

SA24.2 Organizational Management of Telemedicine Technology: Conquering Time & Space Boundaries in Health Care Services

• Olivia R. Liu Sheng; University of Arizona, Dept. of MIS, Tucson, AZ 85721; sheng@bpa.arizona.edu
• Paul Jen-Hwa Hu; University of South Florida, Dept. of Info. Mgmt. & Dec., Coll. of Bus. Admin., Tampa, FL; paulhu@coba.usf.edu
• Chi-Ping Wei; National Sun Yat-Sen University, Dept. of Info. Mgmt., Coll. of Mgmt., Kaohsiung, , Taiwan, ROC;

We address the need for systematic analyses of organizational management of telemedicine. Specifically, concepts of decision making with life-cycle of technologies are combined to investigate important management decisions and their inter-dependencies in organizations adopting such technologies. Important issues across various technologies needed or telemedicine to succeed are indentified and examined.

Vehicle Scheduling & Routing

SA25.1 An Integrated Approach to Bus & Driver Scheduling

• Richard Freling; Erasmus University Rotterdam, Econometric Inst., PO Box 1738, Rotterdam, 3000 DR , The Netherlands; freling@few.eur.nl
• Dennis Huisman; Erasmus University Rotterdam, Econometric Inst., PO Box 1738, Rotterdam, 3000 DR , The Netherlands;
• Albert P. M. Wagelmans; Erasmus University Rotterdam, Econometrics Inst., PO Box 1738, Rotterdam, DR 3000 , The Netherlands; wagelmans@few.eur.nl

Our presentation deals with an integrated approach to bus and driver scheduling. The focus is twofold: we discuss the effectiveness of integration as compared to the traditional sequential approach and we present results for an application to the Rotterdam public transport company in the Netherlands.

SA25.2 A Computer-Aided Solution for the Train Scheduling Problem

• Chueng-Chiu Huang; Ntl. Yunlin University of Science & Tech., 123 University Rd., Sec. 3, Touliu, Yunlin, Taiwan, 640 , ROC; huangcc@pine.yuntech.edu.tw

We analyze how to generate train schedules for the Taiwan Railroad System. Constraints include the arrival and departure timetables for every train at every station which should be generated within a reasonable time, preferably in under 3 minutes. Thus, if a delay should happen for any scheduled train, an updated timetable can be regenerated quickly to ensure safety for the train system.

SA25.3 A Railway Fleet Assignment Model Based on Multi-Commodity Flow

• Mattias Gronkvist; Carmen Systems AB, Odinsgatan 9, Gothenburg, SE-411 03 , Sweden; mattias.gronkvist@carmen.se
• Sami Gabteni; Carmen Systems AB, Odinsgatan 9, Gothenburg, SE-411 03 , Sweden; sami.gabteni@carmen.se

We present a model for solving railway fleet assignment problems based on a multi-commodity flow formulation. The model primarily minimizes the number of vehicles used, but other cost and quality criteria can also be considered. The Swedish Railways use this model in production, and computational results indicate savings of 4-10%.

SA25.4 Scheduling of Transportation of Multiple Products in a Forestry Company

• Celso Carnieri; UFPR, Caixa Postal 19002, Curitiba, PR, 81.531-990, Brazil; carnieri@mat.ufpr.br
• Julio E. Arce; UFPR, Rua Bom Jesus, 650, Curitiba, PR, 80035010 , Brazil; jarce@floresta.ufpr.br
• Jefferson Mendes; Silviconsult, Rua Almirante Tamandare, 1995, Curitiba, PR, 80040010 , Brazil; silviconsult@cwb.conex.com.br

We report a model developed for a large forestry company in Southern Brazil, which needs to schedule the transport of their forestry products from several stands to the customers. The model considers distances, road conditions, types of trucks and loading/unloading queues, aiming the minimization of costs.

Queueing Systems

SA26.1 Expansion Formulae for Characteristics of Soujourn Time in Assembly Systems

• Hayriye Ayhan; Georgia Institute of Technology, Sch. of ISyE, 765 Ferst Dr., Atlanta, GA 30332-0205; hayhan@isye.gatech.edu
• Dong-Won Soe; Georgia Institute of Technology, Sch. of ISyE, 765 Ferst Dr., Atlanta, GA 30332-0205; dongwon@isye.gatech.edu

We consider a class of assembly-type queueing networks with Poisson input process. The dynamics of the assembly system can be captured via a set of difference equations. We develop a Taylor series expansion for the characteristics of stationary sojourn time using these difference equations.

SA26.2 When You Receive the Reward Matters: Implicit Discounting in 'Controlled' Queueing Systems

• Martin L. Puterman; University of British Columbia, Fac. of Commerce, 2053 Main Hall, Vancouver, BC, V6T 1Z2 , Canada; marty@markov.commerce.ubc.ca
• Mark E. Lewis; University of Michigan, Dept. of IOE, 1205 Beal Ave., Ann Arbor, MI 48109-2117; melewis@engin.umich.edu

In sequential decision models, decision-makers should prefer policies that generate rewards sooner. In an admission controlled M/M/1 queueing system, we show that bias optimality is consistent with this preference. To accomplish this, we compare the optimal policies in models in which the gatekeeper receives rewards upon job completion and entry.

SA26.3 Join the Queue with the Shortest Expected Waiting Time: Exact Asymptotics

• Jerome D. Coombs-Reyes; Georgia Institute of Technology, Sch. of ISyE, 765 Ferst Dr., Atlanta, GA 30332-0205; jerome@isye.gatech.edu
• Robert D. Foley; Georgia Institute of Technology, Sch. of ISyE, 765 Ferst Dr., Atlanta, GA 30332-0205; rfoley@isye.gatech.edu

We consider several queues in which the selected arrivals join the queue with the shortest expected waiting time. We obtain exact asymptotics for various rare events such as the probability of having a large number of customers in the system and their distribution among the queues.

SA26.4 Optimal Control of a Nonstationary Queueing System

• Mark E. Lewis; University of Michigan, Dept. of IOE, 1205 Beal Ave., Ann Arbor, MI 48109-2117; melewis@engin.umich.edu
• Hayriye Ayhan; Georgia Institute of Technology, Sch. of ISyE, 765 Ferst Dr., Atlanta, GA 30332-0205; hayhan@isye.gatech.edu
• Robert D. Foley; Georgia Institute of Technology, Sch. of ISyE, 765 Ferst Dr., Atlanta, GA 30332-0205; rfoley@isye.gatech.edu

Consider a finite capacity queueing system with 2 customer classes. Upon arrival, customers offer a reward, which is paid upon acceptance into the system. At known periodic times, the parameters change. Under a conjecture, we show the existence of gain optimal policies that are periodic, monotonic and of threshold form.

Scenario Generation for Simulation Inputs

SA27.1 A Dynamic Method for Controlling Simulation Model Behavior

• Robert Rush; Queues Enforth Development, 432 Columbia St., Ste. 21, Cambridge, MA 02141-1048; rprush@qed.com

We describe a dynamically adaptive TS procedure for controlling the behavioral characteristics of forecasts generated by simulation models used for stochastic planning. The procedure allows decision-makers to perform customized risk analysis by choosing a pattern of future stochasticities for which they want a well-hedged strategy.

SA27.2 A k-Nearest Neighbor Bootstrap for Generating Weather Scenarios

• Balaji Rajagopalan; Intl. Research Institute for Climate Prediction, Lamont-Doherty Earth Observ., PO Box 1000, Rte. 9W, Palisades, NY 10964-8000; rbala@iri.ldgo.columbia.edu
• Upmanu Lall; Utah State University, UWRL, Logan, UT 84322-8200; ulall@c.usu.edu

A method is provided to generate random sequences that honor the statistical properties of historical data. A 'feature vector' is resampled by conditioning on k nearest neighbors of the vector. This is a nonparametric approximation of a multivariate Markov process. The method is applied to generation of daily weather scenarios.

SA27.3 Adapting the Bootstrap to Generate Realistic Simulation Inputs

• Thomas R. Willemain; RPI, Dept. of Dec. Sci./Eng. Syst., 110 8th St., Troy, NY 12180-3590; willet@rpi.edu
• Omer F. Demirel; RPI, 110 8th St., Troy, NY 12180-3590; demiro2@rpi.edu

Built-in input generators make unrealistic assumptions, especially about independence. Alternatively, historical trace data are realistic but provide only a single input stream. The bootstrap can convert one trace into many realistic inputs. We show how to quantify 'realistic' and tune the bootstrap to perform well in this new job.

Tutorial: Fulbright Scholarships: What are They & How to Get One

SA28.1 Fulbright Experience in Poland: A Case Study

• Thaddeus W. Usowicz; California State University, School of Bus., San Francisco, CA 94127-2716; usowicz@sfsu.edu

I will discuss how to obtain a Fulbright scholarship, share my experience and answer questions from the audience. The 'scholarships' are usually for graduate students doing a dissertation. For faculty, there are research awards, combination research/lectureship and 'senior lecturer' awards. The standard award for faculty is for 5 or 10 months...

Operations Management I

SA29.1 Relaxing Channel Separation - Integrating the Virtual Store

• Ralf W. Seifert; Stanford University, Terman Eng. 391, Stanford, CA 94305; seifert@stanford.edu
• Ulrich W. Thonemann; Stanford University, Terman Eng. 336, Dept. of IEEM, Stanford, CA 94305-4024; uthonema@leland.stanford.edu

We characterize the optimal base-stock levels in a distribution system where left-over stock at the retailers can be used to fulfill demand at a virtual store. We quantify the cost savings compared to maintaining strictly separate distribution channels and examine whether dedicated inventory should be used at the virtual store.

SA29.2 Managing Capacity Under Mixed Steady & Random Demand

• Ananth V. Iyer; Purdue University, 1310 Krannert Bldg., Krannert Sch. of Mgmt., West Lafayette, IN 47907; aiyer@mgmt.purdue.edu
• Apurva Jain; University of Washington, Seattle, WA;

We model the use of capacity when demand is a mixture of steady and random orders. We model the lead time and inventory cost impact of the following approaches: pooled capacity, splitting capacity and use of scheduling priorities. We discuss managerial implications of each approach.

SA29.3 Integrated Service Design Research: Issues & Opportunities

• Rohit Verma; DePaul University, Kellstadt Grad. Sch. of Bus., 1 East Jackson Blvd., Chicago, IL 60604; rverma@wppost.depaul.edu,, http://www.depaul.edu/~rverma

While successful marketing offers a product-service package that appeals to the needs and desires of particular segments of customers, this effort is futile without the ability to efficiently manage, control and improve back-end service processes. We provide an overview of research projects in which we attempt to combine market utility models with relevant operating decisions.

SA29.4 Producitivity & Quality of Working Life: Human-Factor-Associated System Variation

• Vivian Yeh; University of Wisconsin, 1513 University Ave., Madison, WI 53706; yehy@cae.wisc.edu
• David A. Nembhard; University of Wisconsin, Dept. of IE, 1513 University Ave., Madison, WI 53706-1572; nembhard@engr.wisc.edu

We investigate the system variation associated with human factors such as work schedules of individual workers, variety of the tasks and allocation of the workers in a manufacturing company. Management and workers benefit from the finding for its implications on the improvement of productivity/efficiency and quality of working life.

Internet & Supply Chain Management

SA30.1 Intelligent Supply Chain Agents

• Mark E. Nissen; Naval Postgraduate School, 555 Dyer Rd., Code SM/Ni, Monterey, CA 93943-5000; mnissen@nps.navy.mil

The integration of buyer and seller supply chain processes is critical for enterprise speed and responsiveness. Intelligent agent technology offers good potential to overcome many limitations of current Web and EDI technologies. We discuss intelligent agents that represent and autonomously conduct business on behalf of various supply chain participants.

SA30.2 The Real-Time Total Private Network Enterprise

• Cosimo Spera; Soma Research & Technology, 3131 Campbell, , CA 95008; cosimo@soma-rt.com

We focus on the theoretical issues and practical challenge that we have faced while implementing part of the real-time private network for a major computer manufacturer. We detail the most relevant phases of such implementation and we report on some of the metrics used to measure the improvements.

SA30.3 Real-World Examples of Internet Enabled Supply Chain Management

• Chris Jones; Aspen Technology, 19204 North Creek Parkway, Bothell, WA 98011;

We discuss some internet enabled supply chain applications that have been deployed or are in development.

SA30.4 Allocation of Capacity over the Internet

• Vipul Agrawal; NYU, Stern Sch. of Bus., 40 West 4th St., Ste. 700, New York, NY 10012; vagrawal@stern.nyu.edu
• Sridhar Seshadri; NYU, Operations Mgmt. Dept., Stern Sch. of Bus., New York, NY 10012; sseshadr@stern.nyu.edu

We compare different web based mechanisms for the pricing and allocation of capacity to customers with heterogeneous preferences. The mechanisms we study include posting of prices, auctions, and resale. We analyze the effect of demand uncertainty on the expected profit of the seller.

Reliability Research

SA31.1 Software Reliability Modeling: A Comparative Discussion

• Jorge L. Romeu; IIT Research Institute, Ctr. for Information Tech., 201 Mill St., Rome, NY 13440; jromeu@iitri.org

Many different software reliability models have been developed and applied in the field. Such models have attempted to assess and forecast this elusive measure of quality, in a manner akin to what was previously done with hardware reliability. We present a comparative discussion of software reliability models, and of their main problems and limitations...

SA31.2 A Proportional Hazard Degradation Model

• Gholamhossein Eghbali; Rutgers University, Dept. of IE, 96 Frelinghuysen Rd., Piscataway, NJ 08854-8018;
• E. A. Elsayed; Rutgers University, Dept. of IE, 96 Frelinghuysen Rd., Piscataway, NJ 08854; elsayed@rci.rutgers.edu

We develop a statistics-based approach for analyzing degradation data obtained at normal operating conditions. The model is a PHDM and it assumes that the degradation criteria at tl and t2 are proportional corresponding values of the degradation path. The degradation paths are nonlinear functions of time and the standard deviation is time-dependent...

SA31.3 System Reliability Estimation with Binomial Sampling & Simulation by Blocks

• Abdullah Konak; Auburn University, Dept. of ISE, Dunstan Hall, Auburn, AL 36849; akonak@eng.auburn.edu
• Alice E. Smith; Auburn University, ISE Dept., 207 Dunston Hall, Auburn, AL 36832; aesmith@eng.auburn.edu

We propose a Monte Carlo procedure for reliability estimation of a system consisting of highly reliable independent components. This procedure requires less random numbers than the crude Monte Carlo. A negative correlation is introduced among the state vectors, which leads to a smaller sampling variance. The efficiency of the procedure dramatically increases with increasing component reliabilities...

SA31.4 Allocation of Developmental Testing Times for Superimposed Nonhomogeneous Subsystem Failure Intensities

• David W. Coit; Rutgers University, Dept. of IE, 96 Frelinghuysen Rd., Piscataway, NJ 08854-8018; coit@rci.rutgers.edu
• Tongdan Jin; Rutgers University, Dept. of IE, 96 Frelinghuysen Rd., Piscataway, NJ 08854-8018;

In the development of new electronic systems, planning of reliability growth test has become more critical and more difficult as testing budgets diminish. A new method is presented to allocate subsystem reliability growth test time to maximize the system mean time between failure or system reliability for limited testing resources...

Telecommunication Topics

SA32.1 Applied Remote Diagnostics in Paper Machinery: Case Valmet Oyj

• Anssi Mattila; University of Jyvaskyla, Sch. of Bus. & Economics, PO Box 35, Asemakatu 4, Jyvaskyla, 40351 , Finland; anmima@tukki.jyu.fi

We evaluate whether the existing technologies in remote diagnostics can be applied to the paper machinery industry. So far, this stream of information technology has been applied, e.g., in medicine. The following issues will be considered: the speed, privacy and creditability of real time data transfer.

SA32.2 Numerical Inversion of Laplace Transform using Piecewise Legendre Polynomials

• Peter W. den Iseger; Erasmus University, Econometric Inst., Rm. H11-29, Fac. of Economics, PO Box 1738, Rotterdam, 3000 DR , The Netherlands; iseger@few.eur.nl

Many important characteristics of models used within the field of OR can be represented in terms of Laplace transforms. With the new method, many important quantities of these models can be calculated with attaining the machine precision. One can also obtain piecewise polynomial approximations of these quantities.

SA32.3 Global Consolidation in the Telecommunications Industry

• Steven R. Powell; California State Polytech University, CIS Dept., 3801 West Temple Ave., Pomona, CA 91768; srpowell@csupomona.edu

Merger and acquisition activity among common carriers in the telecommunications industry has become a worldwide phenomenon. We examine the technological, regulatory, organizational and market factors which are driving the industry's consolidation and address the effect that industry consolidation will have on public telecommunications policy.

Panel: Emerging Issues & Directions in Quality, Statistics & Reliability - Editors' Point of View

SA33.1 Panel: Emerging Issues & Directions in Quality, Statistics & Reliability - Editor's Point of View

• E. A. Elsayed; Rutgers University, Dept. of IE, 96 Frelinghuysen Rd., Piscataway, NJ 08854; elsayed@rci.rutgers.edu
• Hoang Pham; Rutgers University, Dept. of IE, 96 Frelinghuysen Rd., Piscataway, NJ 08854-8018; hopham@rci.rutgers.edu
• Timothy Read; Dupont Company;
• G. Geoffrey Vining; Virginia Tech, Dept. of Statistics, Blacksburg, VA 24061;

This is a special panel to invite journal editors to address their view on the emerging issues & directions in quality, statistics and reliability. Editors from IE Transactions, Journal of Quality Technology, Technometrics and the International Journal of Reliability will contribute. The panel will include short presentations given by each panelist and Q&A discussions.

Achieving Efficient Supply Chain Management

SA34.1 Measures of Quality along a Food Supply Chain

• Jack C. Hayya; Pennsylvania State University, Dept. of MS, 303 Beam Bldg., University Park, PA 16802; jch@psu.edu
• Xin X. He; South Carolina State University, School of Bus., 300 College St. NE, Orangeburg, SC 29117; xhe@scsu.edu

The philosophy of supply chain management involves the elimination of organizational boundaries so that the end-customer can be quickly provided with a safe, fresh product. In this context, we examine how the quality measures evolve as go we from the upstream to the downstream end of the supply chain...

SA34.2 Analysis of Variablility Control Mechanisms in Multi-Echelon Distribution Systems

• Joseph Geunes; University of Florida, Dept. of ISE, 303 Weil Hall, Gainesville, FL 32611; geunes@ise.ufl.edu

Distribution systems often contain several echelons consisting of different firms that interact only through product ordering and delivery. Chain members observe demand and optimize their own operations without considering the impact on system performance. We investigate practical mechanisms that coordinate operations in distribution systems involving manufacturing, warehousing and retail stages.

SA34.3 The Effects of the Shape of Lead-Time Demand Distribution on the Logistics System Costs

• John E. Tyworth; Pennsylvania State University, 509F Bus. Admin. Bldg., Dept. of Bus. Logistics, University Park, PA 16802; jet@psu.edu
• Amy Z. Zeng; University of North Carolina, Cameron Sch. of Bus., Dept. Production & Dec. Sci., Wilmington, NC 28403-3297; zenga@uncwil.edu

Service level or shortage cost is frequently used to deal with the LTD when making inventory-logistics decisions in supply chain management. We clarify the discussion of the appropriate form of the total cost function and examine the sensitivity of the decisions to the shape of the distribution of LTD.

SA34.4 Minimizing Response Time in a Multi-Product Production System with Significant Changeover Times

• Mark L. Spearman; University of Alabama, Coll. of Commerce/Bus. Admin., 3000 Alston Hall, Box 870226, Tuscaloosa, AL 35487-0226; mspearman@cba.ua.edu

We consider the problem of determining the batch sizes that minimize total system flow time for a system that makes several different products in which a significant changeover time occurs whenever the product changes. We obtain the counter intuitive result that smaller process batch sizes should accompany longer changeover times.

Tutorial: Leveraging the Internet in Business Applications

SA35.1 Tutorial: Leveraging the Internet in Business Applications

• Albert Angehrn; INSEAD, Technology Mgmt. Area, Fontainebleau, , France; albert.angehrn@insead.fr

Electronic commerce and Internet-related business applications are an exciting field for research as well as for experimenting with innovative, IT-enhanced learning methods. This tutorial will address the theory and practice of studying and teaching Internet business applications, focusing in particular on issues of high managerial relevance, i.e., linking IT to a company's business strategy...

Data Envelopment Analysis

SA36.1 An Equivalent Model for Solving DEA Models that Excludes the Column(s) being Scored

• Paul P. Rouse; University of Auckland, Dept. of Acct. & Finance, Private Bag, Auckland, 92019 , New Zealand; p.rouse@auckland.ac.nz
• Knox Lovell; University of New South Wales, Dept. of Economics, Sydney, , Australia; k.lovell@unsw.edu.au

DEA models in which the producers being evaluated are excluded from the envelopment LP technology matrix have been proposed for several purposes: as a tie-breaking technique, as a way of identifying outliers, as a way of overcoming the truncation problem when DEA scores are used as depenent variables in second stage regressions...

SA36.2 Indices of Fortune 500 Companies using DEA

• Agha Iqbal Ali; University of Massachusetts, Schoo. of MFOM, Amherst, MA 01003; aiali@som.umass.edu
• Yao Chen; University of Massachusetts, School of MFOM, Amherst, MA 01003; yao@som.umass.edu

We use Fortune 500 rankings that are published annually to produce 3 indices of performance that are computed using DEA. By combining the data in 3 different ways, we define a labor efficiency index, an operation leanness index and a resource utilization index. The interpretation of these indices is illustrated for 2 industries: motor vehicle and computer and office equipment.

SA36.3 Chance Constrained & Cone Dominance in DEA Models

• Zhimin Huang; Adelphi University, Sch. of Mgmt. & Business, 1 South Ave., Box 701, Garden City, NY 11530; huang@panther.adelphi.edu
• Susan Li; Adelphi University, School of Business, 1 South Ave., Box 701, Garden City, NY 11530; susan@panther.adelphi.edu

We present cone chance constrained models in DEA. Efficiency measure of a decision making unit (DMU)is defined via overall cone probabilistic comparisons of inputs and outputs with other DMUs and can be characterized by solving a chance constrained programming problem.

SA36.4 A Unified Additive Model Approach for Evaluating Inefficiency & Congestion with Associated Measures in DEA

• Joe Zhu; Worcester Polytechnic Institute, Dept. of Mgmt., 100 Institute Rd., Worcester, MA 01609; jzhu@wpi.edu
• William W. Cooper; University of Texas, Dept. of MSIS, CBA 4-202, Austin, TX 78712-1174; cooperw@mail.utexas.edu
• Lawrence M. Seiford; National Science Foundation, OR/Production Systems Program, 4201 Wilson Blvd., Arlington, VA 22230; lsieford@nsf.gov

We develop a necessary and sufficient condition for the presence of (input) congestion. Our approach detects congestion, determines the amount of congestion and identifies factors responsible for congestion and distinguishes congestion amounts from other components of inefficiency. These amounts are all obtainable from non-0 slacks in a slightly altered version of the additive model.

Panel: The Art & Tools of Teaching Decision Analysis

SA37.1 Panel: The Art & Tools of Teaching Decision Analysis

• Craig W. Kirkwood; Arizona State University, Dept. of Management, Tempe, AZ 85287-4006; craig.kirkwood@asu.edu
• Gregory S. Parnell; US Military Academy, Dept. of Systems Eng., Mahan Hall, rm. 432, West Point, NY 10996; fg7526@usma.edu
• Jack M. Kloeber, Jr.; Air Force Institute of Technology, Dept. of Operational Sci., 2950 P St., Bldg. 640, WP AFB, OH 45433-7765; jack.kloeber@afit.af.mil
• Michael R. Middleton; University of San Francisco, McLaren Sch. of Business, 2130 Fulton St., San Francisco, CA 94117-1080; middleton@usfca.edu

There have been several excellent texts published in the past five years which, combined with improved software, have greatly increased our ability to teach relevant decision analysis to undergraduate through graduate students. Our panel will discuss issues regarding methods of teaching DA, with comments on innovative teaching styles and techniques and which software works best for different levels of students interested in DA.

Insights from Past Lives of Decision Analysis Practitioners

SB01.1 An Interdisciplinary Look at Decision Analysis

• Todd Anderson; Strategic Decision Group, One Boston Place, 39th Floor, Boston, MA 02108-4467; tanderson@sdg.com

Decision analysis is especially useful in interdisciplinary problems because different effects and constituencies have to be weighted appropriately in order to obtain a solution. This idea will be highlighted by drawing on the author's academic research in biophysical chemistry and industrial experience.

SB01.2 Using Game Theory in Strategic Decision Analysis

• Waseem Noor; Strategic Decision Group, One Boston Place, 39th Floor, Boston, MA 02108-4467; wnoor@sdg.com

For business decision problems with a game theoretic flavor, practical decision analysis tools can be augmented to reintegrate the concepts of decision tree and game tree. We explore when and how this could be useful.

SB01.3 Exploiting Parallelism in Decision Problems

• Robert P. Hewes; Strategic Decision Group, One Boston Place, 39th Floor, Boston, MA 02108-4467; bhewes@sdg.com
• Jeff Keisler; , One Boston Place, 39th Floor, Boston, MA 02108-4467; jkeisler@netzero.net

Many decision problems have an inherently parallel nature, but often times the parallelism goes unexploited. Through analogy with parallel computation, we develop a set of constructs to describe this aspect of decision problems. We demonstrate the concepts by applying them to resource allocation problems.

SB01.4 Ideas for Decision Analysis from Model Theory

• Jeff Keisler; , One Boston Place, 39th Floor, Boston, MA 02108-4467; jkeisler@netzero.net
• H. Jerome Keisler; University of Wisconsin, Dept. of Math, Madison, WI 53706; keisler@uwisc.math.edu

Ideas from model theory, an area within mathematical logic, are explored for potential application in decision analysis. Definitions for concepts such as models, theories, expressability and decidability may be useful in describing how different people in an organization think and what can happen when they come together.

Combinatorial Approaches to Linear Programming Problems

SB02.1 A Combinatorial Optimization Approach to Choosing Population Samples for Genetic Mapping

• Daniel G. Brown; Cornell University, Dept. of Comp. Sci., 4141 Upson Hall, Ithaca, NY 14853; snowman@cs.cornell.edu
• T.J. Vision; Cornell University, Dept. of Plant Breeding, Ithaca, NY 14853;
• S.D. Tanksley; Cornell University, Dept. of Plant Breeding, Ithaca, NY 14853;
• David B. Shmoys; Cornell University, School of OR/IE, Dept. of Comp. Sci., Ithaca, NY 14853;
• R.T. Durrett; Cornell University, Dept. of Math., Ithaca, NY 14853;

Given large populations of test individuals, geneticists seek small samples whose genetic mapping quality is still high. We design LP models for this problem, study its complexity and find samples through randomized rounding and greedy algorithms. While the problem is NP-hard even to approximate, our results are consistently good.

SB02.2 On the Multicommodity-Flow Approach to Low-Congestion Routing

• Aravind Srinivasan; Bell Labs, Lucent Technologies, Rm. 2C-302A, 700 Mountain Ave., Murray Hill, NJ 07974-0636; srin@research.bell-labs.com
• F. Thomson Leighton; MIT, Cambridge, MA;
• Satish Rao; NEC Research Institute;

Multi-commodity flow provides a natural relaxation for several routing problems. We sketch how it is useful in the context of routing in expander graphs: networks that are well known to have good routability and fault-tolerance properties. We also show certain worst-case relationships between frational congestion (a bound that is given by a certain multicommodity flow relaxation) and...

SB02.3 Latest Computational Experiments with Approximate LP-Solver

• Daniel Bienstock; Columbia University, Dept. of IEOR, New York, NY 10027; dano@ieor.columbia.edu

We present ongoing computational experiments involving our implementation of rhe epsilon-approximation approach initially developed by Plotkin, Shmoys & Tardos and Grigoriades & Khachiyan, as applied to the approximate solution of large LPs. On instances of practical interest, our code outperforms commercial LP solvers by a factor of 20 or more.

Supply Chain Management

SB03.1 Optimal Contracting in Long-Term Supplier Relations

• Paul R. Kleindorfer; University of Pennsylvania, The Wharton School, 3620 Locust Walk, Philadelphia, PA 19104-6366; kleindorfer@wharton.upenn.edu
• Moti Levi; University of Pennsylvania, The Wharton Sch., OPIM Dept., 1341 SH-DH, Philadelphia, PA 19104-6366; moti@opim.wharton.upenn.edu

We use a multi-period principal-agent framework to analyze long-term supply relationships in the presence of relationship specific investments. Building on Williamsonian transactions cost theory and case studies, we explore a 'middle way' between markets and hierarchies in which suppliers and buyers make specific investments to enable long-term strategic partnerships.

SB03.2 Competitive Analysis of Capacity Options

• Sridhar Tayur; Carnegie Mellon University, Grad. Sch. of Indust. Admin., Pittsburgh, PA 15213; stayur@grobner.gsia.cmu.edu

A supplier of capacity, such as natural gas storage, provides multiple manufacturers who compete in the same (recently deregulated) market. Motivated by a CEO request to understand strategic issues, we study the effects of uncertainty and number of manufacturers on channel co-ordination, price (hence quantity) and profitability.

SB03.3 Incentive Alignment in a Supply Chain with Channel Assembly & Component Commonality

• Lingxiu Dong; Stanford University, IEEM Dept., Stanford, CA 94305-4024;
• Hau Leung Lee; Stanford University, Dept. of IE/EM, Terman 351, Stanford, CA 94305-4024; haulee@leland.stanford.edu

Due to the risk pooling effect, component commonality is highly desirable to assembly operators. We examine the channel assembly program in the PC industry in which a distributor assembles multiple brand products and orders components from a supplier. We demonstrate that a centralized first best solution in inventory stocking decisions can be achieved...

SB03.4 Informational Effects on Supplier-Retailer Contracting

• Krishnan S. Anand; Northwestern University, KGSM, Sheridan Rd., Leverone Hall, 2001, Evanston, IL 60208-2001; anand@nwu.edu
• Ravi Anupindi; Northwestern University, MEDS Dept., KGSM, 2001 Sheridan Rd., Evanston, IL 60208; r-anupindi@nwu.edu
• Yehuda Bassok; University of Washington, Dept. of MS, Box 353200, Seattle, WA 98195; bassok@u.washington.edu

We develop a dynamic model of a supplier and a retailer, with only the retailer having access to demand information. We show that both informational issues and the retailer's inventory-holding costs are key determinants of the supplier-retailer contract. One consequence is that holding-cost subsidies can be an important strategic weapon for the supplier.

Simulation Applications II

SB04.1 Disease Description: From Worms & Germs to OR/MS

• Chindripu P. Arun; Ayr Hospital, Dept. of Urology, Ayr, KA6 6DX , Scotland, UK; arunpeter@yahoo.com

Just as the germ theory of disease helped describe infectious diseases, the concepts of queueing and inventory are helping describe and understand diseases of pathologic physiology. Previous papers have dealt with queueing models; the present paper describes how the concept of inventory control is advancing our understanding.

SB04.2 Use of OR Tools to Evaluate Improved Bio Response Program Templates

• John D. Walther; US Department of Defense Army, ECBC, E3330, Rm. 133, AMSSB-REN-DA Walther, Aberdeen Proving Gr., MD 21010; jdwalthe@apgea.army.mil
• Freeman F. Marvin; Decision Advantage, PO Box 1467, Middleburg, VA 20118; ffmarvin@hotmail.com
• Susan A. Hayes; Litton-TASC, 4801 Stonecroft Blvd., Chantilly, VA 20151; sahayes@tasc.com

As part of the Congressionally mandated program to improve our nation's response to chem/bio terrorism, a Response Template has been developed to help cities manage activities required to respond to bio incidents. A comprehensive evaluation methodology, using several OR techniques, has been developed to test and validate the template.

SB04.3 A New Efficient Approach to Genetic Algorithms

• Renyou Wang; Carnegie Mellon University, Dept. of Eng. & Public Policy, 5000 Forbes Ave., Pittsburgh, PA 15213; renyou@andrew.cmu.edu
• Urmila Diwekar; Carnegie Mellon University, Dept. of Eng. & Public Policy, 5000 Forbes Ave., Pittsburgh, PA 15213; urmila@cmu.edu

GAs use random mutation and crossover for generation movement. We will incorporate the constrained and stratified sampling techniques, such as Latin hypercube sampling method, median Latin hypercube sampling method and especially Hammersley sampling method, into a GA to improve its performance.

SB04.4 Computer Simulation, the Bullwhip Effect & Long, Serial Assembly Lines

• Thomas F. Brady; Purdue University North Central, 1401 South US Highway 421, Westville, IN 46391; tbradyjr@purduenc.edu

We examine the use of a computer simulation model to analyze a long, complex serial assembly line used in making television sets. In particular, we focus on the identification and analysis of critical areas in the line and show empirical evidence of the bullwhip effect. Finally, we show how optimization was used to improve the performance of the line.

Decision Analysis I

SB05.1 Emerging Roles for Management Science Topics in Strategic Planning

• Abe Feinberg; California State University, MS Dept., MS 8245, Northridge, CA 91330; abe.feinberg@csun.edu,, http://www.csun.edu/~hcmgt004/

In strategic planning, issues in finance, marketing and management often capture the greatest attention. Other than principles of operations management, major roles for MS topics are not obvious. Recent software developments for strategy analysis have included significant uses of DA and simulation. Several examples will be presented.

SB05.2 withdrawn - author request of 10/15

• P. L. Yu; University of Kansas, Sch. of Business, 350E Summerfield Hall, Lawrence, KS 66045-2003; yupl@cc.nctu.edu.tw
• C. I. Chiang; National Chiao Tung University, Inst. of Traffic & Transport., Taipei, Taiwan, , ROC;
• G. H. Tzeng; National Chiao Tung University, Inst. of Traffic & Transport., Taipei, Taiwan, , ROC;

SB05.3 withdrawn - author request of 9/21

• Jacques Teghem; Faculte Polytechnique de Mons, 9 rue de Houdain, Mons, B-7000 , Belgium; teghem@mathro.fpms.ac.be
• Taicir Loukil; University of Sfax, Route de l'aerodrome km4, Sfax, , Tunisia;

SB05.4 An Introduction to Decision Conferencing

• Lawrence D. Phillips; London School of Economics & Political Science, Dept. of Operational Research, Houghton St., London, WC2A 2AE , UK; larry_phillips@msn.com

Decision conferencing is a series of intensive working meetings whose unique feature is the creation, on-the-spot, of a multi-attribute decision analytic model which provides participants with a 'tool for thinking'. The paper presents the latest developments in this 'socio-technical' approach to decision analysis.

Yield Management Applications

SB06.1 A Bid Price Approach to Capacity Expansion in Telecom Services

• Garrett J. van Ryzin; Columbia University, 412 Uris Hall, Grad. School of Bus., New York, NY 10027; gjv1@columbia.edu
• Itir Karaesman; Columbia University, 412 Uris Hall, Graduate School of Bus., New York, NY 10027; ik67@columbia.edu

Telecom service providers must plan the expansion of capacity for inter-office channels (IOCs) and equipment cards based on anticipated demand. Because they have considerable flexibility in provisioning service, determining when and where capacity should be added is complex. We discuss work on a marginal value (bid price) approach to this problem.

SB06.2 Revenue Management with Cancellations, Overbookings & No-Shows

• Shaler Stidham, Jr.; University of North Carolina, Dept. of OR, CB 3180, 210 Smith Bldg., Chapel Hill, NC 27599-3180; sandy@email.unc.edu
• Conrad J. Lautenbacher; NationsBank, NC1-001-04-16, 101 N. Tryon St., Charlotte, NC 28255; conrad.j.lautenbacher@ncmi.com

We report on our continued studies of revenue management with cancellations, overbooking and no-shows, based on formulation as a queueing control problem. New numerical results are presented, along with economic interpretations, approximations and implications for network problems.

SB06.3 Optimal Dynamic Pricing for Perishable Assets with Shifting Customer Reservation Proce

• Wen Zhao; University of Pennsylvania, OPIM Dept., The Wharton Sch., Philadelphia, PA 19104;
• Yu-Sheng Zheng; University of Pennsylvania, The Wharton Sch., OPIM Dept., 1330 SH-DH, Philadelphia, PA 19104-6366;

Price changes are used for 2 reasons: to compensate for statistical demand fluctuations and to adapt to shifts of customers' reservation prices over time. The latter could be more significant. We show the optimal price is decreasing in the inventory, but not necessarily over time if the reservation price shifts.

SB06.4 Managing Rental Services with a Heterogeneous Customer Base

• Sergei Savin; University of Pennsylvania, The Wharton Sch. 1300 SH-DH, 3620 Locust Walk, Philadelphia, PA 19104-6366;
• Morris A. Cohen; University of Pennsylvania, The Wharton School, Dept. of OPIM, Philadelphia, PA 19104-6366; cohen@grace.wharton.upenn.edu
• Noah F. Gans; University of Pennsylvania, OPIM Dept., The Wharton School, Philadelphia, PA 19104-6366; gans@wharton.upenn.edu
• Ziv Katalan; University of Pennsylvania, The Wharton Sch., OPIM Dept., 1314 SH-DH, Philadelphia, PA 19104-6366; katalan@wharton.upenn.edu

Many rental businesses face the problem of managing their service capacity in the presence of both seasonality and demand uncertainty. We develop a modeling framework for this environment that combines long-term, strategic fleet sizing decisions with tactical, yield-management decisions. Structural properties of optimal capacity management policies are derived.

Panel: Research Opportunities & New Initiatives at the National Science Foundation

SB07.1 Panel: Research Opportunities & New Initiatives at the National Science Foundation

• Louis A. Martin-Vega; National Science Foundation, Innovation/Organizational Chng, 4201 Wilson Blvd., Arlington, VA 22230; lmartinv@nsf.gov
• Lawrence M. Seiford; National Science Foundation, OR/Production Systems Program, 4201 Wilson Blvd., Arlington, VA 22230; lsieford@nsf.gov
• Joseph E. Hennessey; National Science Foundation, Innovation/Organizational Chng, 4201 Wilson Blvd., Arlington, VA 22230; jhenness@nsf.gov
• Jeryl L. Mumpower; National Science Foundation, SBER, Rm. 995, 4201 Wilson Blvd., Arlington, VA 22230; jmumpowe@nsf.gov

The National Science Foundation has initiated many research opportunities related to manufacturing, operations research, information technology and logistics. In this panel, these research opportunities and new initiatives at NSF will be discussed.

Improving the Provisioning Process at Lucent Technologies

SB08.1 Applying Target-Costing Methods & Tools to New Products

• H. M. B. Bird; Bell Labs., Lucent Technologies, PO Box 900, Princeton, NJ 08542-0900; hbird@lucent.com
• M. B. Clifton; Bell Labs., Lucent Technologies, PO Box 900, Princeton, NJ 08542-0900; mclifton@lucent.com
• W. P. Townsend; Bell Labs., Lucent Technologies, PO Box 900, Princeton, NJ 08542-0900; wptownsend@lucent.com

Target costing is a market-driven process for developing competitive products. We have applied target costing to a large number of new families of manufactured telecommunications products, with excellent results. We describe our approaches and processes and some of the analytical tools and methods that we have used.

SB08.2 TopOff: A New Pull Manufacturing Link in Factory Supply Chains: Description, Practice & Analysis

• David S. Mohoric; Lucent Technologies, Optical Networking Group, 1600 Osgood St., North Andover, MA 01845-1043; mohoric@lucent.com
• Geoffrey C. Scott; Bell Labs., Lucent Technologies, PO Box 900, Princeton, NJ 08540-0900; gcscott@lucent.com

At the factory level, pull manufacturing is an important link in flexible, responsive corporate supply chains. Recently, a new pull manufacturing paradigm, TopOff, was developed within Lucent. This offers several advantages over traditional pull systems. We describe TopOff, our experiences with implementing it in a large Lucent factory and on-going operations modeling and analysis that promote it.

SB08.3 Risk Analysis helps Lucent Transition to New Technologies

• Kedar S. Naphade; Lucent Technologies, PO Box 900, Princeton, NJ 08542-0900; knaphade@lucent.com

When product technology changes, timing the cut off of materials purchasing for the old generation and the initial purchasing of materials for the new technology is a critical and high dollar risk decision. We describe the problem and a quick and dirty model that assists decision making for such transitions.

Panel: New Directions in Group Decision & Negotiation I

SB09.1 New Directions in Group Decision & Negotiation I

• Melvin F. Shakun; NYU, Stern Sch. of Bus., 44 West 4th St., New York, NY 10012-1126; mshakun@stern.nyu.edu
• Kalyan Chatterjee; Pennsylvania State University, Smeal Coll. of Bus., 310 Beam BAB, University Park, PA 16802; kchatterjee@psu.edu
• Rachel Croson; University of Pennsylvania, Wharton Sch. of Bus., Dept. of OPIM, 1322 SH-DH, Philadelphia, PA 19104-6366; crosonr@wharton.upenn.edu
• Peri H. Iz; Health Care Finance Administration, 7500 Security Blvd., Baltimore, MD 21244; izperi@iamdigex.net
• D. Marc Kilgour; Wilfrid Laurier University, Dept. of Math, Laurier Centre for Military, Waterloo, Ontario, N2L 3C5 , Canada; mkilgour@mach1.wlu.ca
• Brenda Zimmerman; York University, Schulich Sch. of Business, 4700 Keele St., North York, Ontario, M3J 1P3 , Canada; bzimmerm@bus.yorku.ca

This double panel session will discuss new research and application directions in group decision and negotiation. Panel members will make opening remarks followed by panel discussion with audience participation.

Semidefinite Programming & its Applications I

SB10.1 A QQP-Minimization Method for Nonconvex Semidefinite Programs

• Florian Jarre; University of Wuerzberg, Angewandte Mathematik, Am Hubland, Wuerzburg, D-97074 , Germany; jarre@mathematik.uni-wuerzburg.de

We discuss a method for smooth unconstrained nonconvex minimization. Here, the usual linear systems of the Newton step are replaced by simple quadratically constrained quadratic programs. This method is then used within a barrier method for solving nonconvex semidefinite programs. Some global convergence results are given along with numerical examples.

SB10.2 Preconditioned Conjugate Gradient Methods for Large Scale & Dense Linear Systems in Semidefinite Programming

• Kazuhide Nakata; Tokyo Institute of Technology, Dept. of Math. & Comp. Sci., 2-12-1 Oh-Okayama Meguro-ku, Tokyo, 152 , Japan; nakata@is.titech.ac.jp
• Shao-Liang Zhang; Tokyo Institute of Technology, Dept. of Math. & Comp. Sci., 2-12-1 Oh-Okayama Meguro-ku, Tokyo, 152 , Japan; zhang@mmm.t.u-tokyo.ac.jp
• Masakazu Kojima; Tokyo Institute of Technology, Dept. of Math. & Comp. Sci., 2-12-1 Oh-Okayama Meguro-ku, Tokyo, 152 , Japan; kojima@is.titech.ac.jp,, http://www.is.titech.ac.jp/labs/kojimalab/kojima/

To solve large scale and dense linear systems arising in each iteration of primal-dual interior-point method for semidefinite programming, we incorporated the conjugate gradient method into the semidefinite programming algorithm. We present some preconditioners for the conjugate gradient method and their numerical comparison.

SB10.3 A .699-Approximation Algorithm for Max-Bisection

• Yinyu Ye; University of Iowa, Dept. of MS, S384 Pappajohn Bldg., Iowa City, IA 52242-1000; yinyu-ye@uiowa.edu

We present a .699-approximation algorithm for max-bisection, i.e., partitioning the nodes of a weighted graph into two blocks of equal cardinality so as to maximize the weights of crossing edges. This is an improved result from the .651-approximation algorithm of Frieze & Jerrum and the semidefinite programming relaxation of Goemans & Willamson.

SB10.4 Global Optimization of General Nonlinear Programs via Successive Convex Relaxation

• Masakazu Kojima; Tokyo Institute of Technology, Dept. of Math. & Comp. Sci., 2-12-1 Oh-Okayama Meguro-ku, Tokyo, 152 , Japan; kojima@is.titech.ac.jp,, http://www.is.titech.ac.jp/labs/kojimalab/kojima/

We present how we extend the successive convex relaxation methods developed by Kojima and Tuncel for nonconvex quadratic programs to general nonlinear programs.

Verified Global Optimization

SB11.1 Numerica: An Overview & Examples

• Pascal Van Hentenryck; Universite Catholique de Louvain, Dept. of Comp. Science & Eng., 2 Place Sainte-Barbe, Louvain la Neuve, B-1348 , Belgium; pvh@info.ucl.ac.be

No abstract supplied.

SB11.2 An Overview of the GlobSol Package

• R. Baker Kearfott; University of Southwestern Louisiana, Dept. of Math., Box 4-1010, Lafayette, LA 70504-1010; rbk@usl.edu

With hardware and software advances, verified global optimization is becoming increasingly practical. The GlobSol package, initially a research code, has evolved into a general package for rigorous bounds on solutions to a number of practical NLP problems. The capabilities, use and installation of GlobSol will be outlined and illustrated.

SB11.3 Application of the GlobSol Package

• George F. Corliss; Marquette University, Dept. of Math, Stats. & CS, Milwaukee, WI 53233; georgec@mscs.mu.edu

Engineers often must optimize designs of physical or electronic devices by minimizing an objective, e.g., stress or fit to a design curve, that is computed by extensive finite element computations. We apply the 'all-together' method of Dennis & Williamson and use Kearfott's GlobSol to solve the resulting constrained optimization problem.

SB11.4 Compiler & Programming Language Support for Verified Global Optimization

• G. William Walster; Sun Microsystems, MTV12-40, 2550 Garcia Ave., Mountain View, CA 94043; bill.walster@eng.sun.com

Computing verified solutions to constrained and unconstrained nonlinear global optimization problems is one of the most important practical results of interval arithmetic and analysis. Routine practical commercial applications of interval global optimization require both commercial compiler and programming language support of interval data types. The form of this support and the steps required to produce it are described.

Sales Force Models

SB12.1 On the Optimality of Delegating Pricing Authority to the Salesforce

• Kissan Joseph; University of Kansas, 345J Summerfield Hall, Lawrence, KS 66045;

Marketing theory prescribes that delegating pricing authority to the salesforce will improve profits. Surprisingly, the available empirical evidence reveals just the opposite to be true. Thus, we develop a model that re-examines the optimality of the price delegation decision.

SB12.2 Adjusting for Controllable vs Uncontrollable Performance Variables in Performance Evaluation: Effects on Ratings & Fairness

• Cynthia Huffman; University of Delaware;
• Lisa Cain; University of Pennsylvania, The Wharton School, Philadelphia, PA 19104; lisa@marketing.wharton.upenn.edu

We report on 2 experiments examining the effects on fairness and performance ratings of making adjustments for controllable vs. uncontrollable intermediate variables in performance evaluation. Results show that, as expected, performance ratings are adjusted for intermediate variables, here selling price and course difficulty. Results support predictions that performance ratings are more extreme...

SB12.3 Delegating the Promotion Function to the Salesforce under Conditions of Moral Hazard & Adverse Selection

• Sanjog Misra; University of Rochester, Simon Sch. of Business, Rochester, NY;

The issue of moral hazard in the compensation of salespeople has been researched in some detail (Basu et al, 1985, Lal & Srinivasan, 1993). However, most of these models assume that marketing mix elements do not interact with the design of a compensation plan. We discuss the issue of delegating the promotion function to the salesforce...

SB12.4 Compensation Schemes for National Account & Field Representatives

• Pradeep P. Bhardwaj; UCLA, Los Angeles, CA;

Previous research argues that heterogeneity in the salesforce is the main reason why firms offer different contracts to different members of the salesforce. We propose a model to understand why do firms offer different contracts to ex ante homogeneous sales reps?

Measurement & Modeling Issues in Consumer Behavior

SB13.1 The Evolution of Preferences & Choice Behaviors of Consumers in a New Market

• Carrie M. Heilman; Washington University, Olin Sch. of Bus., Box 1133, 1 Brookings Dr., St. Louis, MO 63130;
• Douglas Bowman; Emory University, Goizueta Sch. of Bus., Atlanta, GA 30322; doug_bowman@bus.emory.edu
• Gordon P. Wright; Purdue University, Krannert Grad. Sch. of Mgmt., West Lafayette, IN 47907; gordy@mgmt.purdue.edu

We examine how customer loyalty, price sensitivity and brand preferences evolve over purchases made by consumers who are new to a category. Our theory is based on the notion that the desire to collect information about alternatives and risk aversion to lesser known brands (i.e., private labels) drive initial brand choices. Once information search is completed, subsequent choices reveal loyalty to a preferred brand(s).

SB13.2 When & Why does Consumer Stockpiling Accelerate Consumption Volume?

• Pierre Chandon; London Business School, Sussex Place, Regent's Park, London, NW1 45A , England, UK; pchandon@lbs.ac.uk
• Brian Wansink; University of Illinois, Champaign, IL;

Managers and researchers often assume that each individual has a fixed level of consumption from which different brands can draw. We show how the household stockpiling that can result from sales promotions can accelerate consumption and increase primary demand. By integrating the psychological and economic antecedents of product consumption, a framework is developed...

SB13.3 The Fault is Not in Our Scales but in Ourselves: How Response Styles Weaken Correlations from Surveys

• Eric Greenleaf; NYU, Stern Sch. of Business, 44 West 4th St., Rm. 8-94, New York, NY 10012-1126;
• Geeta Menon; NYU, Stern Sch. of Business, 44 West 4th St., Rm. 8-94, New York, NY 10012;
• Barbara Bickart; Rutgers University, School of Bus.;

We examine how response styles weaken true correlations between marketing variables measured with rating scales, which adds to weakening already caused by scale discreteness. We find empirically that the former often exceeds the latter and assess how it reduces statistical power and distorts market research.

SB13.4 Comparison between Pre- & Post-Purchase Product Evaluation Processes: Exploring Nature of Enduring, Receding & Emerging Evaluative Criteria

• Kimberly Taylor; Florida International University, Mktg. Dept., University Park, Miami, FL 33199;
• Mary Jane Burns; ;

Little previous research specifically links consumers' pre-purchase and post-purchase evaluations. We propose a descriptive framework of the ways in which these 2 sets of criteria might differ, i.e. some types of criteria endure in importance, while others recede or emerge, and test this framework in a laboratory study. Implications of these results for both academic researchers and marketing practitioners are given.

Technology & Organizational Economics

SB14.1 Learning & Leakage: Implications for Alliance Organization

• Joanne E. Oxley; University of Michigan, Sch. of Bus. Admin., 701 Tappan St., Ann Arbor, MI 48109-1234; oxley@o.imap.itd.umich.edu

Inter-firm alliance participants frequently face a dilemma: how do we promote knowledge tranfer and learning while safeguarding valuable intangible assets? We synthesize insights from transaction cost economics and organization theory to explore this, examining relationships between project characteristics, 'knowledge transfer mechanisms' and governance safeguards adopted by alliance participants.

SB14.2 Complementary Capabilities & the Adoption of New Technology: The Case of the Three-Strip Technicolor Process

• Douglas R. Johnson; University of Illinois, 350 Commerce West Bldg., 1206 South 6th St., Champaign, IL 61820; drjohn@cba.uiuc.edu
• Stephen Bowden; University of Waikato, , , New Zealand;

We investigate adoption of the 3-strip technicolor process by directors/cinematographers in the Hollywood studio system. We argue that adoption by these individuals is influenced by their capabilities, their employers' strategic objectives, and most importantly, their access to necessary complementary skills. We use network analysis to assess such access.

SB14.3 Core Technologies, Peripheral Technologies & Corporate Diversification

• Brian S. Silverman; Harvard Business School, Soldiers Field, Boston, MA 02163; bsilverman@hbs.edu

We consider how a firm's technological resources affect its diversification pattern. It integrates elements of TCE and the RBV to explore how a firm's decision to exploit technological resources through diversification, rather than licensing, is affected by the degree to which the technological resources comprise the firm's 'core technology.'

SB14.4 Does it Pay to Let Scientists Do Good Science? Evidence from R&D Job Offers

• Scott Stern; MIT, Sloan Sch. of Mgmt., 50 Memorial Dr., Cambridge, MA 02144;

We evaluate the impact of organizational design on the labor market for scientists/engineers - notably, how 'science-friendly' practices are related to wages. Two theories, 'productivity' and 'preference,' offer conflicting hypotheses concerning this relationship. We introduce a novel field-based approach to evaluate this relationship and consequently distinguish between theories.

International Aspects of Technological Development

SB15.1 Technology Transfer in the International Area: Exploring Technology Transfer Differences

• William T. Flannery; University of Texas, Mgmt. of Technology Program, 6900 North Loop, 1604 West, San Antonio, TX 78249-0634; wflannery@utsa.edu

With the rapid globalization of technology-based economies brought on by advances in IT, international technology transfer is best understood as a dynamic dependent upon geographic and industry specifics. We will examine aspects of the technology transfer process that appear to be interdependent upon these specifics.

SB15.2 Incubating & Sustaining Learning & Innovation Poles in Latin America & the Caribbean

• David L. Gibson; University of Texas, IC2 Institute, 2815 San Gabriel, Austin, TX 78705; davidg@icc.utexas.edu
• Sunil Tankha; University of Texas;
• Julie Nordskog; University of Texas, Austin, TX;
• Jennifer Burtner; University of Texas, Austin, TX;
• Pedro Conceigo; University of Texas, Austin, TX 78705;

We present the logic, conceptual framework and elements of an initial action plan for leveraging codified knowledge and tacit know-how through Web-based networks and face-to-face training programs to accelerate technology-based economic developments in Latin America and the Caribbean.

SB15.3 Contribution of the Technology Asset to the Competitive Advantage in the Chemical Industrial Sectors in a Developing Country

• Rocio Cassaigne-Hernandez; Universidad Nacional Autonoma de Mexico, Fac. of Chemistry, Circuito Institutos Edif D, Mexico, 04510 , Mexico; cassaigl@servidor.unam.mx
• Carlos Escobar-Toledo; Universidad Nacional Autonoma de Mexico, Fac. of Chemistry, Circuito Institutos Edif D, Mexico, 04510 , Mexico;

Technological development is not an easily quantifiable concept. Yet, such determination is necessary in order to identify the stage of development of a technology and to define the starting point for further technological and strategic planning. We present a tested methodology to determine the value of this asset.

SB15.4 Global New Product Development in International Brazilian Firms

• Sergio Takahashi; University of Sao Paulo, Fac. of Economy, Bus. Admin. & Acctg., Ribeirao Preto, 14040-900 , Brazil; sergiota@usp.br
• Vania Passarini Takahashi; University of Sao Paulo, Fac. of Economy, Bus. Admin. & Acctng., Sao Paulo, 14040-900 , Brazil;

We compare the strategies of product development in international firms in Brazil with similar strategies in a global context. This analysis includes aspects of competencies and knowledge management in an organization design perspective and relates to mergers, acquisitions and partnerships.

Finance Applications

SB16.1 An Evaluation of Brazilian Currency Policy

• David Benaroch; , Rua Uruguai 413, Apt. 1208, Rio de Janeiro, 20510-000 , Brazil;
• Silvio R. Badenes de Gouveia; Billinton Metais, Praia de Botafogo 228, Rio de Janeiro, PR, 22359-900 , Brazil;
• Nelson F. Maculan; Federal University of Rio de Janeiro, COPPE-Sistemas, PO Box 68511, Rio de Janeiro, 21945-970 , Brazil; maculan@cos.ufrj.br

We analyze the evolution of the trajectory of the currency policy. An emphasis is adopted with respect to main sectors: financial, production and international trade. A framework of risk is also approached. The foreign investments are mentioned as protection of exchange flotations. Target inflation and Tobing tax are shown as for and against.

SB16.2 Financial Statement Estimation through Entropy

• Thomas R. Jefferson; Sultan Qaboos University, Coll. of Commerce, PO Box 20, Al-Khod, Muscat, 123 , Sultanate of Oman; tjeffers@squ.edu.om
• Waymond Rodgers; University of California, Anderson Grad. School of Mgmt., Riverside, CA; rodgersw@mail.ucr.edu
• Carlton H. Scott; University of California, Grad. Sch. of Mgmt., Irvine, CA 92717; chscott@uci.edu

Entropy is used to find the most probable estimates satisfying accounting relationships. Geometric programming provides a fast solution method. The method estimates actual accounts.

SB16.3 A Priori, Ex Post & Market Variances

• Peter C. Anselmo; New Mexico Tech., Dept. of Mgmt., Box 3, Speare Hall, Socorro, NM 87801; anselmo@nmt.edu
• Erquan Li; New Mexico Tech., Dept. of Comp. Sci., Socorro, NM 87801; erquan@nmt.edu
• James S. Dyer; University of Texas, Dept. of MSIS, Coll. of Bus. Admin., Austin, TX 78712; jim.dyer@bus.utexas.edu

We discuss the connection between a priori risk (usually variance) used in asset allocation and the actual measurement of asset risk via analysis of historical return distributions.

SB16.4 Some Optimization Problems from Time Series Analysis in Finance

• Mustafa Pinar; Bilkent University, Dept. of IE, Ankara, 06533 , Turkey; mustafap@bilkent.edu.tr,, http://pascal.ie.bilkent.edu.tr/~mustafap
• Aslihan Salih-Altay; Bilkent University, Dept. of Mgmt., Ankara, 06533 , Turkey; asalih@bilkent.edu.tr
• Aharon Ben-Tal; Israel Institute of Technology, Faculty of IE & Mgmt., Haifa, 32000 , Israel;

Volatility estimation is an important and challenging problem in finance. We review the autoregressive conditional heteroskedastic variance-covariance model and related maximum likelihood estimation problems. Emphasis is on numerical solution of the resulting optimization models. Preliminary results will be given depending on progress.

Workload Control in Job Shops

SB17.1 Work Load Management under MRP

• Silvanus T. Enns; University of Calgary, Dept. of MME, Fac. of Eng., Calgary, Alberta, T2N 1N4 , Canada; enns@enme.ucalgary.ca

In an MRP environment, planned lead times dictate work-in-process inventory levels while the MPS dictates throughput rates. We investigate how shop load feed back and/or queueing relationships can be used in setting planned lead times to reduce shop inventory while maintaining due date performance and throughput rates.

SB17.2 Workload Control in the Semiconductor Industry

• John W. Fowler; Arizona State University, Dept. of IE, Tempe, AZ 85287-5906; john.fowler@asu.edu
• Gary L. Hogg; Arizona State University, Chandler, AZ 85226-1236;
• Scott J. Mason; Arizona State University, 4714 E. Verbena Dr., Phoenix, AZ 85044;

We describe the state-of-the-art in workload control as applied to the semiconductor industry. The focus is on examining the concepts behind workload control heuristics and evaluating their effectiveness, overhead requirements and implementability. We provide an overview of the distinctive elements in a semiconductor manufacturing environment, general dispatching methods, order release strategies and future directions.

SB17.3 Comparing Work Load Control Theory & Industrial Practice: A Survey in Italy

• R. Cigolini; Polytechnico Milano, Milan, , Italy; roberto.cigolini@polimi.it

Our objective is to understand the reasons for the existing distance between the WLC research and the PP&C procedures used by most companies, and how this gap can be reduced. An empirical survey was carried out and interviews were conducted with 30 shop floor managers. Empirical data were collected and compared...

SB17.4 Robustness of Workload Concepts

• M. Land; University of Groningen; m.j.land@bdk.rug.nl

With the help of controlled release methods, WLCs buffer the shop floor against the dynamics of the incoming flow of jobs. This should provide a lean and stable shop floor environment. To handle a wide spectrum of dynamics, the concepts must guarantee a certain level of robustness...

Network Design & Applications

SB18.1 withdrawn - chair request of 9/21

• Mervat Chouman; Universtie de Montreal, CRT, CP 7128, succ. Centre-ville, Montreal, Quebec, H3C 3J7 , Canada; mervat@crt.umontreal.ca
• Teodor Gabriel Crainic; Universite de Montreal, CRT, CP 7128, succ. Centre-Ville, Montreal, Quebec, H3C 3J7 , Canada; theo@crt.umontreal.ca
• Bernard Gendron; Universite de Montreal, DIRO & CRT, CP 7128, succ. Centre-ville, Montreal, Quebec, H3C 3J7 , Canada; bernard@crt.umontreal.ca

SB18.2 Adaptive Approximation Methods for Dynamic Service Network Design

• Katerina Papadaki; Princeton University, CASTLE Lab., Dept. of Civil Eng. & OR, Princeton, NJ 08544; powell@princeton.edu
• Warren Powell; Princeton University, Dept. of Civil Eng. & OR, Princeton, NJ 08544;

Dynamic service network design problems produce intractably large integer programs. Engineering practice uses simulations with relatively simple dispatch rules. We are developing a class of dynamic programming approximations which produce simulations with optimization behavior which exploit structural results in the value function. Numerical results will be presented.

SB18.3 Airline Network Design

• Cynthia Barnhart; MIT, Ctr. for Transport Studies, Dept. of Civil Eng & OR, Cambridge, MA 02139-4307;
• Andrew P. Armacost; MIT, Ctr. for Transport. Studies, Dept. of Civil Eng. & OR, Cambridge, MA 02139;
• Monoj Lohatepanont; MIT, Ctr. for Transport. Studies, Dept. of Civil Eng. & OR, Cambridge, MA 02139;

We consider the problem of determining the profit maximizing airline flight schedule given a fixed fleet of aircraft. The flight schedule specifies the flight legs to be flown and their associated scheduled departure times. We pose alternative models and algorithms for this problem and present results using data provided by major airlines.

SB18.4 Dual-Ascent Procedures for Multicommodity Capacitated Fixed-Charge Network Design Problems

• Bernard Gendron; Universite de Montreal, DIRO & CRT, CP 7128, succ. Centre-ville, Montreal, Quebec, H3C 3J7 , Canada; bernard@crt.umontreal.ca
• Teodor Gabriel Crainic; Universite de Montreal, CRT, CP 7128, succ. Centre-Ville, Montreal, Quebec, H3C 3J7 , Canada; theo@crt.umontreal.ca
• Antonio Frangioni; Universita di Pisa, Corso Italia 40, Dept. di Informatica, Pisa, 56125 , Italy; frangio@di.unipi.it

We present a dual-ascent approach for computing lower bounds on the optimal value of fixed-chage multicommodity capacitated network design problems. The approach is an extension of previous procedures for particular cases, such as the fixed-charge multicommodity uncapacitated network design problem (Balakrishnan, Magnanti, Wong 1989) and the capacitated plant location problem (Guignard, Spielberg 1977)...

Constraint-Based Scheduling

SB19.1 An Investigation into Two Approaches for Constraint Directed Resource Allocation & Scheduling

• Andrew J. Davenport; University of Toronto, 97 Dovercourt Rd., Toronto, Ontario, M6J 3C2 , Canada; andrewd@eil.utoronto.ca
• J. Christopher Beck; University of Toronto, Enterprise Integration Lab., 4 Taddle Creek Rd., Toronto, Ontario, M5S 3G8 , Canada; chris@eil.utoronto.ca

We investigate new techniques for scheduling with resource alternatives based on the complete decomposition prior to scheduling (CDPS) approach and the multiple alternative decomposition (MAD) approach. We investigate experimentally whether the higher computational cost of decision making in the MAD approach results in better scheduling performance than the CDPS approach.

SB19.2 Incorporating New Scheduling Criteria in Constraint-Based Tools

• Philippe Baptiste; University of Technology of Compiegne, UMR CNRS 6599 HEUDIASYC, Ctr. Recherches de Royallieu, Compiegne Cedex, 60205 , France; philippe.baptiste@hds.utc.fr
• Wim Nuijten; ILOG, 9 rue de Verdun, Gentilly Cedex, 94532 , France; nuijten@ilog.fr

Most of the constraint based tools offer efficient primitives to solve scheduling problems where the goal is to minimize the makespan. Such is not the case for other criteria, e.g., weighted sum of completion times. We propose several research directions that could enable constraint based systems to handle various criteria.

SB19.3 A Constraint Programming Approach for Supply Chain Scheduling

• Mark J. Ringer; JD Edwards World Source Company, 1 Technology Way, U9, Denver, CO 80237; mark_ringer@jdedwards.com

J.D. Edwards, a leading ERP and supply chain software vendor, is developing an advanced planning and scheduling system. A constraint programming approach is used to model, solve and optimize a large and complex manufacturing planning and scheduling problem. We will give an overview of the technical and implementation details encountered using the development of this new system.

SB19.4 Constraint Propagation Techniques for Reservoirs

• Jerome Rogerie; ILOG, 9 rue de Verdun, Gentilly Cedex, 94532 , France; rogerie@ilog.fr
• Wim Nuijten; ILOG, 9 rue de Verdun, Gentilly Cedex, 94532 , France; nuijten@ilog.fr

Reservoirs are resources that allow the definition of both capacity consuming and capacity producing activities on them. Commonly found examples of reservoirs are storage tanks and inventory. We will present some powerful constraint propagation techniques for reservoirs, a subject that despite its practical importance has received relatively little theoretical attention.

Using Bus Location Information in Public Transit

SB20.1 Use of Current Traffice Conditions for Paratransit Dispatching

• Mark Hickman; Texas A&M University, Dept. of Civil Eng., College Station, TX 77843; mhickman@tamu.edu
• Cristina Slabic; Texas A&M University, Dept. of Civil & Environ. Eng., College Station, TX 78843-3136;

We have modified an existing algorithm for the paratransit scheduling problem to account for real-time changes in road traffic conditions. The resulting algorithm can be used to support real-time dispatching decisions. A case study in Houston describes the utility of traffic information in paratransit vehicle dispatching.

SB20.2 Irregularly Sampled Probe Vehicles as a Traffic Sensor

• Daniel J. Dailey; Department of Electrical Engineering, M322 EE/CS Bldg., Box 352500, Seattle, WA 98195; dailey@ee.washington.edu
• C. Elango; Department of Electrical Engineering, M322 EE/CS Bldg., Box 352500, Seattle, WA 98195;

SB20.3 A Markov Chains Model for Bus Arrival Time Estimation with GPS Data

• Wei Lin; Virginia Tech, Dept. of Civil Eng., Blacksburg, VA 24061; whlin@ctr.vt.edu

We observed that delays at neighboring bus stops are correlated. Proper utilization of this correlation information will improve the accuracy of bus arrival time estimation. We discuss the formulation of a Markov chains model for bus arrival time estimation in which the behavior of bus drivers and delays in previous bus stops are incorporated.

SB20.4 A Methodology for Evaluating Real-Time Passenger Information Systems

• Rabi Mishalani; Ohio State University, Dept. of Civil & Environ. Eng., 2070 Neil Ave., 470 Hitchcock, Columbus, OH 43210;
• Mark R. McCord; Ohio State University, Dept. of Civil & Environ. Eng., 470 Hitchcok Hall, 2070 Neil, Columbus, OH 43210-1275;
• Sungjoon Lee; Ohio State University, Dept. of Civil Eng., 470 Hitchcock Hall, 2070 Neil, Columbus, OH 43210-1275;

A probabilistic modeling methodology for evaluating real-time bus arrival information systems is developed. Information is assumed to be provided at bus stops where passengers arrive randomly. Sensitivity to passenger utility functions, operational characteristics, real-time vehicle location data, and bus arrival prediction methods is captured. Analytical and simulation-based results are presented.

Dynamic Traffic Assignment II

SB21.1 Real-Time Schedulability of a Distributed Dynamic Traffic Assignment System

• Nhan Huynh; University of Texas, Dept. of Civ. Eng., ECJ 6.2, Austin, TX 78712;
• Hani S. Mahmassani; University of Texas, Dept. of Civil Eng., ECJ 6.2, Austin, TX 78712-1076;
• Yi-Chang Chiu; University of Texas, Dept. of Civil Eng., ECJ 6.2, Austin, TX 78712;
• Khaled F. Abdelghany; University of Texas, Dept. of Civil Eng., ECJ 6.2, Austin, TX 78712;

A real-time scheduling policy is developed to manage the execution of a distributed DTA system for a large-scale network. The essence of the framework is to ensure the system will meet its required timing and precedence constraints in addition to performing its intended functions. The methodology is validated through use of DYNASMART-X, an established DTA system.

SB21.2 Decomposition Approaches for Dynamic Traffic Assignments

• S. Travis Waller; Northwestern University, Dept. of Civ. Eng., 2145 Sheridan Rd., Tech. Inst., Evanston, IL 60208; travis@trans.civil.nwu.edu
• Yue Irene Li; Northwestern University, Dept. of ISE, 2145 Sheridan Rd., Tech. Inst., Evanston, IL 60208; liyue@nwu.edu
• Athanasis Ziliaskopoulos; Northwestern University, 2145 Sheridan Rd., Evanston, IL 60208; a-z@nwu.edu

Various strategies for DTA are analyzed with regard to the possibility of decomposing them into more manageable sub-problems. Both analytical and heuristic approaches are analyzed. Computational results will be presented and discussed.

SB21.3 A Hybrid A Priori & Online Reoptimization Framework for Real-Time Route Guidance Applications

• Yi-Chang Chiu; University of Texas, Dept. of Civil Eng., ECJ 6.2, Austin, TX 78712;
• Hani S. Mahmassani; University of Texas, Dept. of Civil Eng., ECJ 6.2, Austin, TX 78712-1076;

A hybrid DTA framework that incorporates a priori routing policies and online re-optimization is presented. It introduces a collaborative architecture between the centralized and decentralized DTA controllers which generates robust a priori routing policies and performs quick online re-optimization for real-time route guidance in urban network.

SB21.4 Stable Online Solutions for Route Guidance under Information Systems

• Srinivas Peeta; Purdue University, Sch. of Civil Eng., West Lafayette, IN 47907;
• Ta-Hui Yang; Purdue University, Sch. of Civil Eng., West Lafayette, IN 47907-1284;

Issues vis-a-vis stable regions for on-line traffic assignment solutions are explored. Analysis is performed using simulation and analytic models.

Current Developments in IT Outsourcing

SB22.1 Information Systems Outsourcing Contracts: Theory & Evidence

• Vijay Gurbaxani; University of California, Grad. School of Mgmt., Irvine, CA 92717-3125;

No abstract supplied.

SB22.2 Balance Between Insourcing & Outsourcing: A Survey

• Eric K. Clemons; University of Pennsylvania, The Wharton Sch., 1315 Locust Walk, SH-DH 6366, Philadelphia, PA 19104;
• Lorin Hitt; University of Pennsylvania, The Wharton Sch., 1318 SH-DH, Philadelphia, PA 19104; hitt@wharton.upenn.edu
• Eli M. Snir; University of Pennsylvania, The Wharton Sch., 1348 SH-DH, OPIM Dept., 3620 Locust Walk, Philadelphia, PA 19104; eli@grace.wharton.upenn.edu

No abstract supplied.

SB22.3 The State of the IT Outsourcing Market: A View from the Field

No abstract supplied.

Valuation of Real Options in Competitive Electricity Markets

SB23.1 A Real Option Model for Power Investments

• Yves Smeers; Universite Catholique de Louvain, CORE, 34 Voie du Roman Pays, Louvain la Neuve, 1348 , Belgium; smeers@core.ucl.ac.be
• Kadjo Bedin Kouassi; Universite Catholique de Louvain, CORE, 34 Voie du Roman pays, Louvan La Neuve, 1348 , Belgium;

We consider an investment in a new generation plant when several risk factors are described by affine jump processes. We derive the partial differential equation characterizing the value of the plant and explore the possibilities of solving the associated real option problem by complementary problems.

SB23.2 Exercising Real Operation Options under Price Uncertainty

• Chung-Li Tseng; University of Maryland, Dept. of Civil Engineering, College Park, MD 20742; chungli@eng.umd.edu,, http://www.eng.umd.edu/~chungli
• Graydon L. Barz; Stanford University, Dept. of EES & OR, Stanford, CA 94305; gbarz@leland.stanford.edu

A power plant converts a particular fuel to electricity. This conversion process involves 2 commodities with different market prices. The operation of a power plant associated with intertemporal constraints is formulated as real options. We present optimal exercising strategy for the real option at each time period.

SB23.3 Dispatchability under Uncertainty: A Real Options Approach

• Todd Strauss; PHB Hagler Bailly Inc., 1776 Eye St. NW, Washington, DC 20006; tstrauss@haglerbailly.com

Operating flexibility at an electric power plant is valued as a real option. The real option approach is contrasterd with the standard discounted cash flow approach and with a pure financial option valuation approach.

SB23.4 Incorporating Physical Constraints & Startup Costs in Option-Based Valuation of Generation Capacity

• Shijie Deng; Georgia Institute of Technology, School of ISyE, Atlanta, GA 30332-0205; deng@ieor.berkeley.edu
• Shmuel S. Oren; University of California, Dept. of IE/OR, 4172 Etcheverry Hall, Berkeley, CA 94720-1777; oren@ieor.berkeley.edu

We describe a stochastic DP approach for valuing electricity generation capacity with physical constraints and startup costs. Generators are modeled as a strip of cross commodity call options with a delay and cost imposed on each option exercise. Stochastic prices of electricity and fuel are represented by a recombing trinomial tree.

Management Issues in Health Care Technology Management I

SB24.1 Organizational Integration, Product Development Proficiency & Market Success in the Medical Device Industry

• Murray Millson; Syracuse University, Innovation Mgmt. Prog., Sch. of Mgmt., Syracuse, NY 13244;
• David Wilemon; Syracuse University, Innovation Mgmt. Prog., Sch. of Mgmt., Syracuse, NY 13244; dwilemon@som.syr.edu

What are the relationships between organization integration, product development process proficiency and market success for medical device products? Based on an empirical investigation, we examine these relationships and offer several managerial and research implications for those concerned with product development in the medical device industry.

SB24.2 Women in Medicine: Perceptions of Technological Competence

• Pamela L. Cox; Cazenovia College, 304 Eddy Hall, Cazenovia, NY 13035; pcox@cazcollege.edu

As more women enter a profession, decreases in salary, prestige and promotional opportunities often follow. One of the reasons these changes seem to occur is that perceptions of technological competence tend to decrease with the increase in the proportion of women. We examine the perceptions of technological competence associated with women in medicine.

SB24.3 Defining & Managing Error Reduction Technologies in the Blood Transfusion Processes

• John R. Grout; Berry College, Campbell Sch. of Bus., Mount Berry, GA 30149; jgrout@campbell.berry.edu

Transfusion operations require matching individual-specific materials and tasks with patients. Human errors occur relatively frequently. Jigs, fixtures and more advanced technologies are possible counter-measures. Extended fault tree analysis is used to provide a set theoretic definition of mistake-proofing and a system for managing mistake-proofing so process efficiency is not impaired.

SB24.4 Bio-Medical Models: A Theoretical Leap for Management Science

• Paula E. Bobrowski; SUNY, Dept. of Mktg., Oswego, NY 13126; bobrowsk@oswego.edu
• Arun Ranchod; McKesson Automated Health Care, Product Line;

According to Drucker, current management models have become outdated in an information age. The information revolution is redefining the business enterprise as a creator of value and wealth based on innovation. We demonstrate how information theory has been mathematically modeled and applied in the field of biomedical research and illustrate how these theories can be applied in creating management models...

Routing

SB25.1 Solving Vehicle Routing Problems with Soft Time Windows

• Robert A. Russell; University of Tulsa, Finance & OM Dept., 600 South College Ave., Tulsa, OK 74104-3189; rrussell@utulsa.edu
• Wen-Chyuan Chiang; University of Tulsa, Finance & OM Dept., 600 South College Ave., Tulsa, OK 74104-3189; wen-chyuan-chiang@utulsa.edu

We examines VRPs with soft time windows. We compare the effectiveness of classic ejection chain techniques with a TS metaheuristic employing advanced recovery strategies. We also explore the applicability of smoothing transformations to time window problems.

SB25.2 A Heuristic Approach to the Period Vehicle Routing Problem

• Burcin Bozkaya; ESRI, 380 New York St., Redlands, CA 92373; bbozkaya@esri.com
• Buyang Cao; ESRI, 380 New York St., Redlands, CA 92373; bcao@esri.com

We present a model and a heuristic algorithm for solving the period vehicle routing problem. We develop a two-phase cluster-and-schedule where we implement TS in both phases. We integrate this algorithm with a client-server DSS and develop visit schedules for a major elevator manufacturer.

SB25.3 Capacity-Sharing Strategies for Stochastic Vehicle Routing Problems

• Alan L. Erera; University of California, 109 McLaughlin Hall #1720, Berkeley, CA 94720-1720; alerera@ieor.berkeley.edu

A new family of operating strategies for single-period stochastic VRPs based on capacity sharing is proposed and preliminary analysis is reported. The estimation of expected operating costs and near-optimal configuration of systems is discussed.

SB25.4 withdrawn - author request of 10/8

• Anulark Pinnoi; Asian Institute of Technology, ISE Program, PO Box 4, Klongluang, Patumtani, 12120 , Thailand; anulark@ait.ac.th
• Ha Dang Son; Asian Institute of Technology, ISE Program, PO Box 4, Klongluang, Patumtani, 12120 , Thailand;

Fluid & Queueing Networks

SB26.1 Stochastic Networks with Quality of Service Specifications

• Costis Maglaras; Columbia University, Columbia Business Sch., 409 Uris Hall, 3022 Broadway, New York, NY 10027-6902; c.maglaras@columbia.edu
• Jan A. Van Mieghem; Northwestern University, Kellogg Grad. Sch. of Mgmt., 2001 Sheridan Rd., Evanston, IL 60208-2001;

We consider multiclass queueing networks that offer probabilistic QoS guarantees on end-to-end delay and blocking that are differentiated for each exogenous arrival stream. We model these constraints using fluid approximations and then proceed to describe scheduling and admission control policies that guarantee the desired specifications.

SB26.2 Bounds for Multiclass Queueing Networks: From Stability to Performance

• Dimitris Bertsimas; MIT, Sloan Sch. of Mgmt., 50 Memorial Dr., Bldg. E53-363, Cambridge, MA 02142; dbertsim@aris.mit.edu
• David Gamarnik; IBM, TJ Watson Research Ctr., PO Box 218, Yorktown Heights, NY 10598; gamarnik@watson.ibm.com
• John N. Tsitsiklis; MIT, Dept. of Electrical Eng., 77 Massachusetts Ave., 35-209, Cambridge, MA 02139; jnt@mit.edu

Given a piecewise linear Liapunov function that is a witness of stability for a particular policy in a multiclass queueing network, we derive explicit upper and lower bounds on the performance of this policy.

SB26.3 A Necessary & Sufficient Condition for Stability of Fluid Networks

• Hengqing Ye; HKUST, Dept. of IE/EM, Clear Water Bay, Kowloon, , Hong Kong; ieyehq@stu.ust.hk
• Hong Chen; HKUST, Dept. of IE/EM, Clear Water Bay, Kowloon, , Hong Kong; hongchen@uxmail.ust.hk

We study the stability of fluid networks, which implies the stability of queueing networks. We prove that a necessary and sufficient condition for the stability of a fluid network (under either a priority service discipline or a FIFO discipline) is that there exists a Lyapunov function for its fluid level processes...

SB26.4 On the Stability Gap between Fluid & Queueing Networks

• John J. Hasenbein; University of Texas, OR Group, Dept. of Mechanical Eng., Austin, TX 78712; jhas@mail.utexas.edu
• Jim G. Dai; Georgia Institute of Technology, Sch. of ISyE, 765 Ferst Dr., Atlanta, GA 30332-0205; dai@isye.gatech.edu
• John Vande Vate; Georgia Institute of Technology, Sch. of ISyE, 765 Ferst Dr., Atlanta, GA 30332-0205; jvandeva@isye.gatech.edu

In earlier research, Dai & VandeVate demonstrated conditions that completely characterize the global stability region for 2-station multi-type fluid networks. An outstanding question is the relationship between such fluid networks and their queueing counterparts. We demonstrate that the stability region under static buffer priority disciplines may depend on the entire distributions of the service and arrival processes...

Perturbation Analysis & Simulation I

SB27.1 Perturbation Analysis via Coupling

• Liyi Dai; Washington University, Dept. of Systems Sci. & Math., CB 1040, St. Louis, MO 63130; liyi@zach.wustl.edu

Coupling is a method of generating multiple random samples. We are concerned with perturbation analysis via coupling. This approach offers a great versatility of the form of gradient estimators, which is potentially useful for variance reduction and for efficient implementation. Several known perturbation analysis techniques can be reviewed as special ways of coupling...

SB27.2 A Comparison of Perturbation Analysis Techniques

• Michael C. Fu; University of Maryland, Smith School of Bus., Van Munching Hall, College Park, MD 20742-1815; mfu@rhsmith.umd.edu
• Jian-Qiang Hu; Boston University, Dept. of Mfg. Eng., 15 Saint Mary's St., Boston, MA 02215;

We will give an overview on several different perturbation analysis techniques and their relationships, as well as their connection to other derivative estimation methods, such as the likelihood ratio method.

SB27.3 Perturbation Analysis for Stochastic Fluid Queueing Systems

• Weibo Gong; University of Massachusetts, Dept. of Elec. & Comp. Eng., Amherst, MA 01003; gong@golf.ecs.umass.edu;, http://www-unix.ecs.umass.edu/~gong
• Yong Liu; University of Massachusetts, Dept. of Elect. & Comp. Eng., Amherst, MA 01003; yonliu@golf.ecs.umass.edu
• Yujing Wu; University of Massachusetts, Dept. of Elect. & Comp. Eng., Amherst, MA 01003; yuwu@golf.ecs.umass.edu

We discuss perturbation analysis and related rational approximation results for stochastic fluid queueing systems and their application in communication networks.

Mathematical Applications Applied to a Social Agenda

SB28.1 Simulating Group Responses to Non-Traditional Survey Instruments

• K. Scott Alberts; St. Olaf College, Dept. of Mathematics, 1520 St. Olaf Ave., Northfield, MN 55057; alberts@stolaf.edu

While statistical simulation is now routine in many areas, social scientists still avoid such techniques. During a larger study, a measure of item agreement was developed as part of a forced-distribution Q-sort instrument. This measure was sufficiently complicated that simulation was used to get reliable estimates for the moments.

SB28.2 Portfolio Selection: A Comparison between Cluster Analysis & the Analytic Hierarchy Process

• Rafael Solis; California State University, Dept. of IS/DS, Craig Sch. of Bus., Fresno, CA 93720; rafaels@csufresno.edu

Portfolio selection based on historical performance of stocks is conducted utilizing 2 quantitative techniques: cluster analysis and AHP. The portfolios created using the 2 methodologies contain stocks that have performed differently so as to minimize the overall risk (the same idea as diversification). The 2 portfolios will be observed for 6 months prior to INFORMS Philadelphia...

Operations Management II

SB29.1 Contracting Over Multiple Parameters: Capacity Allocation in Semiconductor Manufacturing

• Suman Mallik; General Electric Company, Bldg. K1, Rm. 5B32A, PO Box 8, Schenectady, NY 12301; suman.mallik@crd.ge.com

Motivated by the experiences of a major US-based semiconductor manufacturer, we will present an integrated model of incentive problems arising in forecasting and capacity allocation. The model involves multiple product managers and multiple manufacturing managers. This is a generalization of our previous work. The summary of models and results of both the works will be discussed.

SB29.2 Multi-Period Capacity Allocation with Machine Duplication in Semiconductor Manufacturing

• Bulent Catay; University of Florida, DIS Dept., 351 Stuzin Hall, Gainesville, FL 32611; catay@grove.ufl.edu
• Selcuk S. Erenguc; University of Florida, DIS Dept., 351 Stuzin Hall, Gainesville, FL 32611-7169; erenguc@notes.cba.ufl.edu
• Asoo J. Vakharia; University of Florida, DIS Dept., 351 Stuzin Hall, Gainesville, FL 32611; vakhariaj@notes.cba.ufl.edu

We address the problem of assigning batches of wafers to different machines by recognizing capacity limitations of the individual machines as well as reducing operating and investment costs related to the machines and inventory holding costs. A mixed-integer mathematical model is proposed and lower bounding and heuristic methods are developed.

SB29.3 Production Planning & Order Assignment for Space Launch Vehicles

• Leslie O. Morgan; University of Utah, 1645 Campus Ctr. Dr. #106, Salt Lake City, UT 84112-9304; mgtlm@business.utah.edu
• Alysse R. Morton; University of Utah, 1645 Campus Ctr. Dr. #106, Salt Lake City, UT 84112-9304; mgtarm@business.utah.edu
• Richard L. Daniels; Georgia Institute of Technology, DuPree Coll. of Mgmt., 755 Ferst Dr., Atlanta, GA 30332-0520; rich.daniels@mgt.gatech.edu

An aerospace firm produces launch vehicles to shoot payloads of different sizes into space. Vehicles can be built in varying sizes, each with known capacity and cost. We consider the problem of jointly determining the optimal mix of launch vehicles to produce and the optimal assignment of satellites to vehicles.

SB29.4 Using Aggregate Estimation Models for Order Acceptance in a Decentralized Production Control Structure for Batch Chemical Mfg.

• Jan C. Fransoo; Eindhoven University of Technology, PO Box 513, Pav F12, Eindhoven, NL-5600 MB, The Netherlands; j.c.fransoo@tue.nl,, http://www.tm.tue.nl/vakgr/lbs/fransoo.htm
• Wenny Raaymakers; Eindhoven University of Technology, PO Box 513, Pav F12, Eindhoven, 5600 MB , The Netherlands;
• J. W. M. Bertrand; Eindhoven University of Technology, PO Box 513, Eindhoven, 01370 , The Netherlands;

We explore the performance of various aggregate models for order acceptance and loading in no-wait job shops. Simulation results show that a linear regression-based model outperforms a workload-based model on capacity utilization and decentralized replanning effort, specifically under increased capacity utilization and/or high variety in the job mix.

Agents, Brokers & Intermediaries

SB30.1 Designing Artificial Agents for Electronic Commerce: Results & Prospects

• James Laing; University of Pennsylvania, Wharton Sch. of Bus., Dept. of OPIM, 1322 SH-DH, Philadelphia, PA 19104-6366; laing@wharton.upenn.edu
• Steven Kimbrough; University of Pennsylvania, The Wharton School, Philadelphia, PA 19104-6366; kimbrough@wharton.upenn.edu

We address design of adaptive artifical agents for automated negotiations. Such agents would be useful in electronic commerce. Our previous work evolved populations of negotiation strategies that rival human performance. The new results reported here apply to designing single (i.e., deployable) agents to negotiate in these difficult contexts.

SB30.2 Evolving the Electronic Market with Trusted Third Parties

• Sulin Ba; University of Southern California, Marshall Sch. of Business, Los Angeles, CA 90089-1421; sulin@usc.edu
• Andrew B. Whinston; University of Texas, Coll. of Bus. Admin., CBA 5.202, Austin, TX 78712-1175; abw@utexas.edu
• Han Zhang; ;

Fear of security breaches and mistrust in the products offered online remain major obstacles to the growth of electronic commerce. The issue of trust has arisen recently with the creation of TTP. Will people use their services? How does the electronic market evolve? The paper studies the role of TTPs in the electronic market.

SB30.3 Designing Behaviors for Information Agents

• Keith S. Decker; University of Delaware, Dept. of Computer & IS, 103 Smith Hall, Newark, DE 19716; decker@cis.udel.edu,, http://www.cis.udel.edu/~decker/

We will discuss the behaviors and architecture of WWW-based autonomous software agents that gather information in groups for humans and other computational agents. We will also report on organizational 'middle-agents' for use in such systems. Examples are drawn from the WARREN project for open multi-agent financial portfolio management, and our current DECAF agent-building toolkit.

SB30.4 On Designing a Multi-Agent System for Distributed Information Filtering

• Rajeev Raje; Indiana University, Dept of CIS, 723 West Michigan St., SL 280, Indianapolis, IN 46202-5132; rraje@cs.iupui.edu
• Snehasis Mukhopadhyay; Indiana University, Dept of CIS, 723 West Michigan St., SL 280, Indianapolis, IN 46202-5132; smukhopa@cs.iupui.edu
• Mathew Palakal; Indiana University, Dept of CIS, 723 West Michigan St., SL 280, Indianapolis, IN 46202-5132; mpalakal@cs.iupui.edu
• Javed Mostafa; Indiana University, Sch. of Library & Info. Sci., Bloomington, IN 47405-1801; jm@indiana.edu

Due to heterogeneity and geographical distribution, the design of global-scale systems is a major challenge. Such systems can be efficiently realized using cooperation between smart, autonomous objects (or agents). We will discuss the design of a multi-agent distributed information filtering system, an example of a global-scale information system.

Quality & Reliability Engineering Applications

SB31.1 Integrating PID Controllers into Statistical Process Control Applications

• Melike Baykal-Gursoy; Rutgers University, Dept. of IE, 96 Frelinghuysen Rd., Piscataway, NJ 08854; gursoy@rci.rutgers.edu
• Edward T. Nugent; Rutgers University, Dept. of IE, 96 Frelinghuysen Rd., Piscataway, NJ 08854;

We consider pure-gain and first-order processes subject to white noise disturbances and random mean shifts. These processes are monitored using SPC techniques. We demonstrate that performance improvements could be achieved using off-the-shelf PID controllers. The addition of PID control requires modification of the statistical process monitoring scheme. We consider three types of mean shifts...

SB31.2 Design of Accelerated Life Testing Plans with Multiple Stresses

• E. A. Elsayed; Rutgers University, Dept. of IE, 96 Frelinghuysen Rd., Piscataway, NJ 08854; elsayed@rci.rutgers.edu
• Lixia Jiao; Rutgers University, Dept. of IE, 96 Frelinghuysen Rd., Piscataway, NJ 08854;

ALT is a widely used approach that estimates reliability of components or systems at normal operating conditions using data obtained at accelerated conditions. The accuracy of reliability estimates depends on the stress levels and the corresponding sample sizes. We present an efficient and practical approach for determining the appropriate stress levels...

SB31.3 Bringing Quality Awareness to Policing

• Alok Baveja; Rutgers University, School of Bus., Camden, NJ 08102; baveja@crab.rutgers.edu
• Michael A. Redmond; Rutgers University, Dept. of Computer Sci., Camden, NJ 08102; redmond@crab.rutgers.edu

Much use of technology for law enforcement has been for operational and tactical purposes. We argue quality awareness can be created using proactive tools that foster relationships among enforcement agencies, community groups and city officials. We present examples of how such an awareness-centered police department can be better equipped to take on the challenges of law enforcement in the future.

SB31.4 Analysis of Purchasing Contracts for an Imperfect Process with Time-Dependent Product Value

• Linguo Gong; Louisiana State University, Dept. of ISDS, Baton Rouge, LA 70803-6316;
• Kwei Tang; Louisiana State Univ., Dept. of ISDS, Baton Rouge, LA 70803;
• Jen Tang; Purdue University, Krannert Sch. of Mgmt., Rm. 421, West Lafayette, IN 47907-1310;

We study the contractual relationship in a single period manufacturing and distribution supply chain when the value of the product decreases over time and the manufacturer's process is subject to random failure. Two contractual relationships are considered. In the first, the 2 parties share the revenue...

Distributed Computing & Network Design Problems in Telecommunications

SB32.1 Exact Algorithms for Tree Topologies in Telecom Network Design

• Trilochan Sastry; Indian Institute of Management, Vastrapur, Ahmedabad, 380015 , India; sastry@iimahd.ernet.in

We study exact algorithms for several variations of the capacitated network design problem, all of which are known to be NP-hard. We describe a tree enumeration algorithm and a dynamic programming path-based algorithm. The worst case running times of the 2 is faster for problem instances where the capacity constraint on edges is very tight...

SB32.2 Examining the Diffusion of Distributed Database Technology

• Balaji Rajagopalan; Illinois State University, 5520 Dept. of Acctng., Normal, IL 61790-5520; brajag@ilstu.edu
• Rathindra Sarathy; Illinois State University, Dept. of Acct. & IS, 5520 College of Bus., Normal, IL 61790-5520; rsarathy@ilstu.edu

Our objective is to use established diffusion models to study the extent of DDB technology in organizations. Swanson's (1994) tri-core model of IS innovation, as well as research by Grover et al. (1997) will be used in the study of DDB diffusion.

SB32.3 Client Server Network Design: Formulations & Algorithms

• Sabyasachi Mitra; George Institute of Technology, DuPree Sch. of Mgmt., Atlanta, GA 30332-0520; saby.mitra@mgt.gatech.edu
• Sridhar Narasimhan; George Institute of Technology, DuPree Sch. of Mgmt., Atlanta, GA 30332-0520; sri.narasimhan@mgt.gatech.edu

We will focus on the design of 3-tier router based client server networks using current communicatin technologies such as ATM and frame relay. Several alternate formulations of the problem with varying levels of complexity will be discussed. Solution methodologies and computational results will also be discussed.

Panel: Quality Improvement & Reliability - Emerging Trends

SB33.1 Panel: Quality Improvement & Reliability - Emerging Trends

• Douglas C. Montgomery; Arizona State University, Dept. of IMSE, Tucson, AZ 85721; doug.montgomery@asu.edu
• Soren Bisgaard; University of Wisconsin, Ctr. for Qual. & Prod. Improv., Dept. of IE, 610 Walnut St., Madison, WI 53705-2397; bisgaard@engr.wisc.edu
• Way Kuo; Texas A&M University, Dept. of IE, College Station, TX 77843; way@acs.tamu.edu

This is a panel discussion by distinguished researchers in the areas of quality and reliability. The panel will discuss emerging trends shaping current and future research in quality improvement and reliability. The panel will include presentations and discussions by leading researchers.

Tutorial: Electronic Business Process Optimization: Enabling a New Era of Business Intelligence

SB34.1 Tutorial: Electronic Business Process Optimization: Enabling a New Era of Business Intelligence

• Irfan M. Ovacik; i2 Technologies, 909 East Las Colinas Blvd., 16th Fl., Irving, TX 75039; ovacik@i2.com
• Willie Weng; i2 Technologies, 909 East Las Colinas Blvd., Ste. 1600, Irving, TX 75063; willie_weng@i2.com

eBPO is a new initiative launched by i2 Technologies to provide decision support solutions that optimize and integrate core business processes across a company while intelligently connecting the enterprise to its trading partners. We are going to focus on supply chain management, which is one of the eBPO processes.

Tutorial: Industry Clockspeed

SB35.1 Tutorial: Industry Clockspeed

• Charles H. Fine; MIT, Sloan Sch. of Mgmt., 30 Wadsworth St., E53-390, Cambridge, MA 02139; charley@mit.edu

No abstract supplied.

DEA I

SB36.1 DEA with Reverse Inputs: Measuring the Relative Efficiencies of Major League Baseball Teams

• Herbert F. Lewis; SUNY, Harriman Sch. of Mgmt & Policy, Stony Brook, NY 11794-3775; hlewis@notes.cc.sunysb.edu
• Thomas R. Sexton; SUNY, Harriman Sch. of Mgmt & Policy, Stony Brook, NY 11794-3775; tsexton@notes.cc.sunysb.edu

We develop a model in which measurement scales for certain inputs are such that larger values indicate lower input consumption. Properties and interpretations of this model are discussed. The model is applied to measure the relative efficiencies of major league baseball teams.

SB36.2 Intertemporal Productivity Analysis & Model Invariance in DEA

• Anthony D. Ross; Texas A&M University, Mays Coll. of Business, College Station, TX 77840; aross@acs.tamu.edu

We present an approach for identifying best practice decision making units in a multiproduct supply network operation. Using incomplete operational data presents many challenges when used by researchers and analysts. An invariant DEA model and some procedures for handling such empirical realities are specified. The model examines the productive efficiency of DMUs over multiple planning horizons.

SB36.3 On Returns to Scale under Weight Restrictions in DEA

• Kaoru Tone; National Graduate Institute for Policy Studies, 2-2 Wakamatsu-cho, Shinjuku-ku, Tokyo, 162-8677 , Japan; tone@grips.ac.jp

We extend the concept of returns to scale in DEA to the weight restriction environments. The definition and computational procedure will be presented. It is demonstrated that the most productive scale size will usually narrowed by this addition of weight restrictions. An empirical study will follow.

SB36.4 Overcoming the Inherent Dependency of Data Envelopment Analysis Efficiency Scores: A Bootstrap Approach

• Mei Xue; University of Pennsylvania, Dept. of OPIM, The Wharton Sch., Philadelphia, PA 19104-6366; xuem@wharton.upenn.edu
• Patrick T. Harker; University of Pennsylvania, Dept. of OPIM, The Wharton Sch., Philadelphia, PA 19104-6366; harker@wharton.upenn.edu,, http://opim.wharton.upenn.edu/~harker

The efficiency scores generated by DEA models are clearly dependent on each other in the statistical sense. Hence, the conventional procedure of regression analysis generally followed in the DEA literature is invalid. We provide a bootstrap method to overcome this dependency problem.

Panel: Integrative Cases for Teaching OR/MS in the MBA Curricula

SB37.1 Panel: Integrative Cases for Teaching OR/MS in the MBA Curricula

• Salwa Ammar; Le Moyne College, Dept. of Bus. Admin., Syracuse, NY 13214; ammars@palm.lemoyne.edu
• Kenneth Baker; Dartmouth College, Tuck School of Bus. Admin., 100 Tuck Hall, Hanover, NH 03755; kenneth.r.baker@dartmouth.edu
• Dana R. Clyman; University of Virginia, Darden Grad Sch. of Bus. Adm., Charlottesville, VA 22906-6550; clymand@darden.gbus.virginia.edu
• Stephen G. Powell; Dartmouth College, Tuck Sch. of Bus. Admin., 100 Tuck Hall, Hanover, NH 03755; sgp@dartmouth.edu
• Janet M. Wagner; University of Massachusetts, Dept. of MSIS, 100 Morrissey Blvd., Boston, MA 02125; profwagner@aol.com,, wagner@umbsky.cc.umb.edu
• Wayne L. Winston; Indiana University, Kelley School of Bus., Bloomington, IN 47405; winston@indiana.edu
• Ronald H. Wright; Le Moyne College, Dept. of Bus. Admin., Syracuse, NY 13214-1399; wrightr@palm.lemoyne.edu

Participants will discuss and share cases which support the learning of OR/MS concepts and another area in the MBA curriculum. Applications include OR/MS and economics, organizational behavior, MIS and more. Discussion to follow on ways to foster integration within the MBA curriculum.

Asking the Right Questions: Experiences Teaching & Learning about Decision Analysis Practice

SC01.1 Developing Corporate Decision Analysis Capability

• Marcy B. Conn; Strategic Decisions Group, 2440 Sand Hill Rd., Menlo Park, CA 94025; mconn@sdg.com

Experience with hundreds of executives and professionals provides insight into the following aspects of teaching decision analysis in a corporate setting: Which concepts are readily transferable in a classroom setting? What skills require substantial hands-on experience and coaching? What are typical timetables for various levels of skill acquisition?

SC01.2 Building Decision Analysis Capability: An Instructional Framework Tested by Student Experience

• Stephen L. Derby; , 2075 Tasso St., Palo Alto, CA 94301; sgderby@pacbell.net
• David G. Lowell; Santa Clara University, OMIS Dept., 323 Kenna Hall, Santa Clara, CA 95053; davelowell@earthlink.net

Decision analysis learned in a classroom does not prepare the student to conduct a decision analysis in practice. We discuss a framework for teaching the practice of decision analysis that evolved from student experiences in a projects course at Stanford's EES&OR Department.

SC01.3 Teaching & Learning about Decision Analysis Practice

• David G. Lowell; Santa Clara University, OMIS Dept., 323 Kenna Hall, Santa Clara, CA 95053; davelowell@earthlink.net
• Stephen L. Derby; , 2075 Tasso St., Palo Alto, CA 94301; sgderby@pacbell.net

In our experience teaching decision analysis practice to executives and graduate students in business and engineering, we have found the most important lessons are listening well, communicating clearly, modeling concisely and taking an iterative approach. We discuss approaches to facilitate students learning these skills.

Approximation Algorithms for Problems in Combinatorial Optimization

SC02.1 Approximation Algorithms for Facility Location Problems

• Fabian Chudak; IBM TJ Watson Research Center, Rt. 134, Yorktown Heights, NY 10598; chudak@watson.ibm.com
• David B. Shmoys; Cornell University, School of OR/IE, Dept. of Comp. Sci., Ithaca, NY 14853;
• David Williamson; IBM TJ Watson Research Ctr., TJ Watson Research Ctr., PO Box 218, Yorktown Heights, NY 10598;
• Karen Aardal; Utrecht University, Houston, TX 77005-1892;

We describe some recent results on approximation algorithms for several variants of facility location problems.

SC02.2 withdrawn - chair request of 8/30

• Kamal Jain; Georgia Institute of Technology, 330492 Georgia Tech Station, Atlanta, GA 30332-1325; kjain@cc.gatech.edu
• Vijay Vazirani; Georgia Institute of Technology, Coll. of Computing, Atlanta, GA 30332;

SC02.3 Approximation Algorithms for a Directed Network Design Problem

• Vardges Melkonian; Cornell University, 290 Rhodes Hall, Ithaca, NY 14853; vardges@orie.cornell.edu
• Eva Tardos; Cornell University, 290 Rhodes Hall, Ithaca, NY 14853;

We consider the problem of finding a minimum cost subgraph in a digraph such that for each vertex set S there are at least f(S) arcs leaving S. A 2-approximation algorithm is presented for the case when f(S) is crossing supermodular. We also describe the basic solutions of the LP-relaxation.

Coordination Issues in Design & Development

SC03.1 The Role of Suppliers in Product Development: An Empirical Investigation in the Auto Industry

• Tim Laseter; University of Virginia, Darden Grad. Sch. of Bus., PO Box 6550, Charlottesville, VA 22906-6550;
• Kamalini Ramdas; University of Virginia, Darden School, 100 Darden Blvd., Charlottesville, VA 22903; ramsask@darden.gbus.virginia.edu

We develop a classification of automobile components and of different types of supplier roles in product development. We then investigate the relationship between the nature of specific components and the role played by their suppliers in product development, through an empirical study in the automobile industry.

SC03.2 Managing the Marketing-Manufacturing Interface during Production Ramp-Up

• Christian Terwiesch; University of Pennsylvania, The Wharton Sch., Dept. of OPIM, SH-DH 1300, Philadelphia, PA 19104; terwiesch@wharton.upenn.edu
• Teck-Hua Ho; University of Pennsylvania, The Wharton Sch., Mktg. Dept., 1400 SH-DH, Philadelphia, PA 19104;

In many high-tech industries, learning and process improvement during production ramp-up are driven by engineering experimentation rather than operator experience. Experiments lead to process improvements, but they also reduce capacity in the short run. We present a formal model of this trade-off raising interesting problems at the marketing-manufacturing interface.

SC03.3 Understanding the Communication Process in Product Development Organizations

• Manuel E. Sosa; MIT, Sloan Sch. of Mgmt., Cambridge, MA 02139; msosa@mit.edu
• Steven D. Eppinger; MIT, Sloan School of Mgmt., Rm. E53-347, Cambridge, MA 02139; eppinger@mit.edu
• Michael T. Pich; INSEAD, Boulevard de Constance, Fontainebleau Cedex, 77305 , France; micheal.pich@insead.fr

To better understand how product development teams communicate, we propose a framework consisting of 4 elements: communication drivers, communication barriers, communication media, and message. We will present several hypotheses and empirical evidence to explore how these 4 elements are related.

SC03.4 Mediating Marketing, Manufacturing & Supply Chain Decisions during Product Development

• Karl Ulrich; University of Pennsylvania, The Wharton School, 1317 SH-DH, Philadelphia, PA 19104; ulrich@wharton.upenn.edu

During new product development, decisions about features and market segments must be made. These decisions are linked to decisions about supply chain configuration, product architecture, and component commonality. This talk reports on progress in developing decision support tools to mediate these decisions.

Price Formations in Auctions

SC04.1 High Bids & Broke Winners

• Charles Zheng; Northwestern University, Dept. of Economics, 2003 Sheridan Rd., Evanston, IL 60208; czheng@nwu.edu

We provide a new auction theory where bidders have budget constraints and they can default. We precisely predict equilibrium bidding behavior and explain the 'high bids and broke winners' anomaly in the C-block FCC spectrum auction. We also advise the seller on how to raise his expected profit by appropriately subsidizing the bidders.

SC04.2 Asymmetric Mergers in Second-Price Common Value Auction Problems

• Vladimir Mares; Rutgers University, Dept. of Economics, New Brunswick, NJ 08903-5062; mares@economics.rutgers.edu

We will identify the equilibria in asymmetric common value auctions. Results contradict the common intuitions about the anti-competitive effects of mergers and joint bidding. We identify an informational externality that leads to a higher confidence in the private signals, alleviates the winner's curse and allows bidders to pursue more aggressive bidding strategies and increase revenues.

SC04.3 Low Revenue Equilibria in Simultaneous Ascending Price Auctions

• Richard Engelbrecht-Wiggans; University of Illinois, Dept. of Bus. Admin., 1206 South 6th St., Champaign, IL 61820; epuls17@uiuc.edu
• Charles M. Kahn; University of Illinois, Champaign, IL;

The outcomes of some recent FCC auctions look similar to those that one might expect if the bidders colluded. We show that such outcomes can be generated as perfect equilibria of simultaneous ascending-price auctions without relying on future auctions, signals, or outside agreements to sustain collusion.

Decision Analysis II

SC05.1 The Beneficiary Factor & Consistency on Decision-Making

• Pedro J. Steiner Neto; CEPPAD-UFPR/FAE, Rua Padre Anchieta, 1231 Ap. 31, Curitiba, PR, 80730-000 , Brazil; pedro.steiner@avalon.sul.com.br
• Luciane G. Silva; CEPPAD-UFPR/FAE, R. Dr. Faivre, 405-6 Andar, Curitiba, PR, 8006-0000 , Brazil;
• Paulo H. Prado; CEPPAD-UFPR, R. Dr. Faivre, 405-6 Andar, Curitiba PR, 80060000 , Brazil;

According to prospect theory, decision makers don't have a consistent behavior, as opposed to rationality theories like the expected utility. We discuss a new reason for such inconsistency, the decision beneficiary. Depending on who will benefit from the decision - capital market, product market or organization - the behavior will differ. Furthermore we identify more consistency for older decision makers and for those occupying top positions as well.

SC05.2 Decision Analysis & the Operationalization of the Precautionary Principle

• Ahti A. Salo; helsinki University of Technology, Systems Analysis Lab., PO Box 1100, Espoo, 02015 , Finland; ahti.salo@hut.fi,, http://www.sal.hut.fi

The precautionary principle provides a guideline for managing risks associated with considerable scientific uncertainties, e.g., GMOs. Drawing on a recent joint project for the European Commission, we suggest that the management of such risks can be supported by considering several decision analytic framings and strategic qualitative factors such as reversibility and resilience.

SC05.3 Using Mathematical Models in Negotiation & Auction Processes

• Rafael Epstein; University of Chile, Dept. of IE, PO Box 2777, Republica 701, Santiago, , Chile; repstein@dii.uchile.cl
• Jaime Catalan; University of Chile, Dept. of IE, PO Box 2777, Republica 701, Santiago, , Chile; jcatalan@dii.uchile.edu

We show how mathematical models help to successfully implement auction processes. We stress quantitative and qualitative benefits of these tools during the negotiation. We show 2 experiences, an auction process for providing meals to half of the public schools in Chile and a negotiation process to provide transportation service to a private pulp company.

SC05.4 A Measure of Decision Making Performance Grounded in Information Physics

• Scott T. Serich; George Washington University, 1603 Kirby Rd., McLean, VA 22101-3207; serich@acm.org

Traditional micro-economic models can be plagued by market imperfections that preclude reliable demand or factor forecasts, particularly in information-intensive industries such as software or consulting. We introduce a new measure of managerial decision-making performance based on optimal energy utilization and leveraging recent theoretical developments in information physics.

Supply Chain Economics

SC06.1 Modeling Information Flows in Supply Chains

• Krishnan S. Anand; Northwestern University, KGSM, Sheridan Rd., Leverone Hall, 2001, Evanston, IL 60208-2001; anand@nwu.edu

We argue that comprehensive models of supply chains need to incorporate both material and information flows and their interaction. The key attributes of ISs are illustrated in the context of a specific model. Using inventories as an example, we show that different informational assumptions could lead to very different operational implications.

SC06.2 Information Economics: The Value of Learning in a Supply Chain

• Dawn Barnes-Schuster; University of Chicago, Kellogg Grad. Sch. of Bus., 1101 East 58th St., Chicago, IL 60637; dawn.barnes-schuster@gsb.uchicago.edu

Using the constructs of IT, we consider the problem of capacity reservation and pricing in a simple 2-stage supply chain. We provide a mechanism for valuing market information under this scenario and relate it to expected system profits.

SC06.3 Multiple Source Procurement Arrangements

• Wedad Elmaghraby; NYU, Stern School of Bus., New York, NY 10012; welmaghr@stern.nyu.edu

As companies find it beneficial to outsource certain parts of their production process, the optimal types of outsourcing arrangements must be considered. The benefits of multiple sourcing vs. sole sourcing are considered under various market conditions.

SC06.4 Restricted Capacity & Choosing Breadth of Distribution

• Martin Lariviere; Duke University, Fuqua Sch. of Business, Durham, NC 27708; larivier@mail.duke.edu
• Gerard Cachon; Duke University, Fuqua Sch. of Business, Box 90120, Durham, NC 27708-0120;

A supplier with limited capacity may simply decide not to sell to some retailers. While potentially lowering sales, limiting distribution allows for better service to participating retailers. We consider how restricting distribution affects supply chain performance and retailer market actions.

Scheduling, Logistics & Facilities Design I

SC07.1 A Column Generation Approach for Integrated Scheduling & Logistics of Assemblies

• Rakesh Nagi; SUNY, Dept. of IE, 342 Bell Hall, Buffalo, NY 14260; nagi@eng.buffalo.edu
• Muhammad Farooq Anwar; SUNY, Dept. of IE, 342 Bell Hall, Buffalo, NY 14260; anwar@eng.buffalo.edu

We address the integrated problem of scheduling machines and logistic activities to minimize the production makespan of complex assemblies. For industrial-sized problems, a column generation approach is developed. The master problem selects from a set of choices, the best sequence of operations, while the subproblem generates new beneficial sequences.

SC07.2 Evaluating Rescheduling Strategies

• Jeffrey W. Herrmann; University of Maryland, Inst. for Systems Research, College Park, MD 20742; jwh2@eng.umd.edu
• Guilherme E. Vieira; University of Maryland, Inst. of Systems Research, College Park, MD 20742; gevieiera@eng.umd.edu
• Edward Lin; University of Maryland, Inst. of Systems Research, College Park, MD 20742; lin@eng.umd.edu

Simulation-based factory scheduling is a powerful tool. However, it must be used properly to attain maximum effectiveness, especially in a dynamic environment that requires rescheduling. We will discuss the influence of rescheduling methods on the performance of discrete-parts manufacturing systems.

SC07.3 A Linear Model for Scheduling Commercial Advertisements on Cable TV

• Jayant Rajgopal; University of Pittsburgh, Dept. of IE, 1048 Benedum Hall, Pittsburgh, PA 15261; rajgopal@engrng.pitt.edu,, http://www.pitt.edu/~rajgopal

We overview the dynamics of TV commercial scheduling and introduce the notion of an audience loss function. A simple, yet robust linear model is described for optimizing this loss function, subject to the constraints that are typically inherent in the process. A numerical illustration is also provided.

SC07.4 AMPOWER: Advanced Manpower Planning System

• Mehmet Gulsen; United Airlines, World Headquarters WHQKB, PO Box 66100, Chicago, IL 66100; mehmet.gulsen@ual.com

AMPOWER is a combined forecasting/optimization system that provides optimal pilot staffing for both short term operational needs and long term strategically planning needs of the airline. The model initially generates forecasts for various absenteeism factors, i.e. sick leave, military duty, etc., that contribute to manpower loss...

Supply Chain Management in the Semiconductor Industry

SC08.1 Supply Chain Management in the Semiconductor Industry: A Software Supplier Perspective

• Y. S. Chang; i2 Technologies USA, 303 Twin Dolphin Rd., Ste. 210, Redwood City, CA 94065; ys_chang@i2.com

Given the complexity of a typical supply chain, supply chain software systems should help companies to intelligently manage the activities of the supply chain. The solution provided by this system enables each company to identify and focus on key business drivers that determine success in its particular portion of the industry. We introduce a new industry-focused supply chain solution.

SC08.2 Supply Chain Management in the Semiconductor Industry: An Equipment Supplier Perspective

• David Anderson; SEMI/SEMATECH, 2706 Montopolis Dr., Austin, TX 78748; dave.anderson@asu.edu

The semiconductor industry is projected to double in size by the year 2005. It is important that the value chain infrastructure (suppliers of equipment, materials, and services) grow along with the semiconductor manufacturers as the industry expands. Developing a strong infrastructure of suppliers is necessary to achieve and maintain manufacturing technology leadership...

SC08.3 Supply Chain Management in the Semiconductor Industry: An IC Manufacturer Perspective

• John Konopka; IBM, Raleigh, NC;

The semiconductor industry is projected to grow to an estimated 300 billion dollars (US) by the year 2005. This paper presents a review of the theories and strategies that have been developed and implemented through various industry wide initiatives to establish globally competitive semiconductor supply chains.

SC08.4 Supply Chain Management in the Semiconductor Industry: Survey Results

• John W. Fowler; Arizona State University, Dept. of IE, Tempe, AZ 85287-5906; john.fowler@asu.edu

We will share results from a recent survey of SEMATECH and SEMI/SEMATECH member companies to gain information on supply chain management within the semiconductor industry. The survey specifically asks about customer relationships, supplier relationships and IT. The survey was to develop a better understanding of customer and supplier relationships within the semiconductor industry...

Panel: New Directions in Group Decision & Negotiation II

SC09.1 New Directions in Group Decision & Negotiation II

• Melvin F. Shakun; NYU, Stern Sch. of Bus., 44 West 4th St., New York, NY 10012-1126; mshakun@stern.nyu.edu
• Kalyan Chatterjee; Pennsylvania State University, Smeal Coll. of Bus., 310 Beam BAB, University Park, PA 16802; kchatterjee@psu.edu
• Rachel Croson; University of Pennsylvania, Wharton Sch. of Bus., Dept. of OPIM, 1322 SH-DH, Philadelphia, PA 19104-6366; crosonr@wharton.upenn.edu
• Peri H. Iz; Health Care Finance Administration, 7500 Security Blvd., Baltimore, MD 21244; izperi@iamdigex.net
• D. Marc Kilgour; Wilfrid Laurier University, Dept. of Math, Laurier Centre for Military, Waterloo, Ontario, N2L 3C5 , Canada; mkilgour@mach1.wlu.ca
• Brenda Zimmerman; York University, Schulich Sch. of Business, 4700 Keele St., North York, Ontario, M3J 1P3 , Canada; bzimmerm@bus.yorku.ca

This double panel session will discuss new research and application directions in group decision and negotiation. Panel members will make opening remarks followed by panel discussion with audience participation.

Eigenvalue Optimization

SC10.1 Semidefinite Relaxation of the Hartree-Fock Approximation in Quantum Chemistry

• Francois Oustry; INRIA Rhone Alpes, ZIRST, 655 av de l'Europe, Montbonnot St Martin, 38330 , France; francois.oustry@inria.fr,, francois.oustry@inrialpes.fr
• Bastiaan J. Braams; New York University, Courant Inst. of Math. Sci., 251 Mercer St., New York, NY 10012; braams@cims.nyu.edu
• Michael L. Overton; New York University, Courant Inst. of Math. Sci., 251 Mercer St., New York, NY 10012; overton@cs.nyu.edu
• Madhu Nayakkankuppam; Simon Fraser University, Pacific Inst. for Math. Sci., Ctr. Exper./Construct. Math, Burnaby, BC, V5A 1S6 , Canada; madhu@cs.nyu.edu

The most common and important variational technique used in many-particle calculations on fermions systems is the Hartee-Fock Approximation. We start here from a quadratic formulation given in the space of density matrices by Garrod & Percus (1954). Then we use Lagrangian duality to obtain a semidefinite relaxation of the problem.

SC10.2 Semistable Programming

• Michael L. Overton; New York University, Courant Inst. of Math. Sci., 251 Mercer St., New York, NY 10012; overton@cs.nyu.edu
• James V. Burke; University of Washington, Box 354350, Dept. of Math., Seattle, WA 98195; burke@math.washington.edu

In semistable programming, the key constraint is that the eigenvalues of a matrix must lie on or to the right of the imaginary axis, generalizing semidefinite programming to non-Hermitian matrices. We give a necessary condition for local minimizers, and give a generalization of the complementarity condition familiar from semidefinite programming.

SC10.3 Optimizing Matrix Stability

• Adrian Lewis; University of Waterloo, Dept. of Combinatorics & Opt., Fac. of Mathematics, Waterloo, Ontario, N2L 3G1 , Canada; aslewis@orion.math.uwaterloo.ca
• James V. Burke; University of Washington, Box 354350, Dept. of Math., Seattle, WA 98195; burke@math.washington.edu
• Michael L. Overton; New York University, Courant Inst. of Math. Sci., 251 Mercer St., New York, NY 10012; overton@cs.nyu.edu

A variation on semistable programming is minimizing the spectral abscissa (largest real part of the eigenvalues) over a given affine set of matrices. We show that the optimal solution often has a large Jordan block, making even small problems quite challenging. The spectral abscissa determines system stability in control theory.

Recent Developments in Global Optimization

SC11.1 Supervised Clustering using Genetic Algorithms

• Ayhan Demiriz; RPI, Decision & Eng. Systems Dept., 110 8th St., Troy, NY 12180-3590; demira@rpi.edu
• Kristin P. Bennett; RPI, Math Sciences Dept., 110 8th St., Troy, NY 12180; bennek@rpi.edu

In supervised clustering, the objective function of an unsupervised technique, k-means clustering, is modified to minimize both the within cluster variance of the input attributes and a measure of cluster impurity based on the class labels. A GA optimizes this complex objective function to produce clusters. Strong experimental results support the approach.

SC11.2 Feature Selection in Clustering

• W. Nick Street; University of Iowa, MS Dept., 108 Pappajohn Bldg., Iowa City, IA 52242; nick-street@uiowa.edu

In many unsupervised learning applications it is important not only to find good clusters but also to identify the subset of predictive features that most strongly influences the cluster formation. We present a mathematical programming formulation for this task, and apply it to disease diagnosis from genetic markers.

SC11.3 Portfolio Selection with Skewness Preference

• Laila Haerian; Towson University, Dept. of Math, 8000 York Rd., Towson, MD 21252-0001; lhaeri1@tiger.towson.edu
• Hong Seo Ryoo; Towson University, Dept. of Math, 8000 York Rd., Towson, MD 21252-0001; hryoo@towson.edu

Markowitz' seminal mean-variance model for portfolio selection is based upon the assumption that an investor's utility can be well-approximated by the first 2 moments of the rate of the return of the portfolio. It is well known, however, that the rate of stocks usually does not satisfy normality assumption...

SC11.4 DC Programming, Global Optimality Conditions & Global Optimality Algorithms

• Alexander Strekalovsky; Irkutsk State University, Lab. of Global Opt. Methods, Lermentov str. 134, Irkutsk, 6640033 , Russia; strekal@icc.ru

No abstract supplied.

Marketing on the Internet

SC12.1 Product Evaluations on the Internet

• Steven M. Shugan; University of Florida, 2030 NW 24th Ave., Gainesville, FL 32605; shugan@dale.cba.ufl.edu
• Larry Winer; University of Florida, 2030 NW 24th Ave., Gainesville, FL 32605;

With the growth of new channels of distribution, such as the Internet, buyers enjoy a wide range of both products and product information. Some suppliers of product information and evaluations, however, have potential conflicts of interest that may cause biased evaluations. We provide an empirical analysis showing that a bias does exist...

SC12.2 Mass Customized Services on the Internet: An Empirical Approach for Pricing & Demand Assessment

• John Liechty; University of Michigan;
• Venkatram Ramaswamy; University of Michigan, Grad. Sch. of Business, 701 Tappan St., Ann Arbor, MI 48109;
• Steven Cohen; Stratford Associates;

We discuss a menu-based, Bayesian approach for assessing customers' price sensitivities and preferences for the variety of features that can be offered by a firm engaged in mass customization. The proposed approach provides valuable guidance in setting prices for a customized portfolio of features and services, so as to increase customer value and revenues...

SC12.3 An Approach to Identify Effective Locations for the Placement of Banner Ads

• Amit Bhatnagar; University of Wisconsin, Milwaukee, WI; amit@uwm.edu
• Purushottam Papatla; University of Wisconsin, Milwaukee, WI; papatla@csd.uwm.edu

Advertising on the internet involves placing banners and/or buttons at other web sites. Unlike traditional advertising that is pushed onto consumers, however, internet ads have to pull consumers in. It is, therefore, important for online businesses to precisely target their ads at locations are likely to attract the type of customers who might be interested in the message...

Product Assortments

SC13.1 How Variety in Product Assortment can Affect Consumption Quantity

• Barbara E. Kahn; University of Pennsylvania, The Wharton School, Dept. of Mktg., Philadelphia, PA 19104;
• Brian Wansink; University of Illinois, Champaign, IL;

We examine the influence of variety on consumption. It is hypothesized that, up to a limit, larger sets will result in more items consumed, as consumers may feel that they want to 'complete the set,' or reach closure by trying all available items. Experiments conducted on 1st and 4th grade students support the hypotheses. Other experiments conducted on adults identify some limiting conditions when the effects are not likely to hold.

SC13.2 Consumer Strategies for Filling the Shopping Basket

• Joel Steckel; New York University, Mgmt. Ed. Ctr., Stern School, 44 West 4th St., New York, NY 10012-1126;
• Jack K. H. Lee; ;

Different uses, users and usage occasions for a given product category give rise to the possibility of multiple preference contexts within individuals and households. Each preference context is assumed to have a unique preference structure, hence, utility for a given alternative. Context dependence drives the need for consumers to fill their 'shopping baskets' with assortments of goods...

SC13.3 The 'Combo Meal' Effect: How to Increase Variety without Really Trying

• Andrea Morales; University of Pennsylvania, The Wharton School, Philadelphia, PA 19104;
• Barbara E. Kahn; University of Pennsylvania, The Wharton School, Dept. of Mktg., Philadelphia, PA 19104;

We present experimental results that indicate that if items already present a menu are repeated in the form of a special 'combo meal,' the perceived variety of the menu, and ultimately the satisfaction with the choice process and the desire to return to the restaurant, are increased. This increase in perceived variety only occurs if the 'combo meal' is priced differently (higher/lower) than the sum of the individual items in the combination.

Coping with Complexity & Interdependence

SC14.1 Too Much of a Good Thing? Product Proliferation & Organizational Failure

• William Barnett; Stanford University, Grad. Sch. of Bus., L-306, Palo Alto, CA 94305-5015; barnett_william@gsb.stanford.edu
• John Freeman; University of California, Haas Sch. of Bus., S545, Berkeley, CA 94720-1900; freeman@haas.berkeley.edu

When organizations make important changes, such as introducing new products, they may gain strategic advantages but also experience disruptions. We argue that these disruptions are especially strong when organizations introduce multiple products simultaneously. To test this hypothesis, we study the effects of new product introductions on the survival of US semiconductor manufacturers.

SC14.2 The Virtue of Simplicity: Navigating Complex Environments with Practices, Norms & Heuristics that Balance Change & Stability

• Anjali Sastry; University of Michigan, Business School, 701 Tappan St., Ann Arbor, MI 48109-1234; masastry@umich.edu
• Fiona Lee; University of Michigan, Business School, 701 Tappan St., Ann Arbor, MI 48109-1234; fionalee@umich.edu

To investigate how successful organizations pair change in some domains with stability in others, we investigate the lifetime evolution of the Financial Women's Association, a non-profit organization for Wall Street women. FWA operations in a volatile environment but its mission, culture and leadership approach have persisted. Interdependent rules and practices maintain continuity while allowing some contradiction and change.

SC14.3 Organizational Decay: How Innovation Stresses Organizations

• Michael W. Lawless; Duke University, Fuqua Sch. of Bus., Durham, NC 27708-0120; lawless@mail.duke.edu
• Nancy A. Staudenmayer; Duke University, Fuqua Sch. of Bus., Box 90120, Durham, NC 27708-0120; nstauden@mail.duke.edu

When new technology generations emerge faster than older decline, strategies to pace technology and cut development cycles may return less profit than expected because relationships among products, units and functions become more complex. Organizational decay is the condition where marginal costs of adopting technology increase with technology generations and interdependence.

SC14.4 Not Only by Strategic Partnerships Alone: Managing a Portfolio of Relationships

• Ben Bensaou; Harvard University, Dept. of Bus. Admin., Soldiers Field, Boston, MA 02163; bbensaou@hbs.edu

Using original data on 447 supplier relationships in the US and Japanese auto industries, we examine whether strategic partnerships are the best way to manage supply. Contrary to popular belief, auto manufacturers in both countries seem to rather manage a portfolio of relationships determined by technological and market factors.

Technology & Innovation Management I

SC15.1 Performance in New Technology Introductions: A Comparison of Firm Structure

• Jeffrey T. Macher; University of California, F534 Haas Sch. of Business, Berkeley, CA 94720; macher@haas.berkeley.edu,, http://www.haas.berkeley.edu/~macher

We examine the introduction of new manufacturing processes into semiconductor firms' wafer fabrication facilities. Two research questions are posed: Are performance differences discernable by firm structure? Can intrafirm capabilities explain these differences? Two measures of performance are tested: rate of improvement in manufacturing yield following introduction of a new process and the length of time of new process development...

SC15.2 Managerial Antecedents of Technological Opportunism in Firm Response to Technologies

• Raji Srinivasan; Pennsylvania State University, 701 CBAB, University Park, PA 16802; rxs213@psu.edu

Firms, sometimes even within a given industry, respond differentially to technology developments. We propose that the extent to which organizations differ in their ability to exercise strategic choice is reflected in their technological opportunism: defined as the ability of a firm to sense technological developments in the environment and to proactively respond to these developments to further its business objectives...

SC15.3 A Model to Guide Product Platform Development for a Global Value Chain

• B. Joon Park; Georgia Institute of Technology, DuPree Coll. of Mgmt., 755 Ferst Dr., Atlanta, GA 30332-0520; gt7104d@prism.gatech.edu
• Soumen Ghosh; Georgia Institute of Technology, Dupree Sch. of Mgmt., 755 Ferst Dr., Atlanta, GA 30332-0520; soumen.ghosh@mgt.gatech.edu
• Nagesh N. Murthy; Georgia Institute of Technology, DuPree Sch. of Mgmt., 755 Ferst Dr., Atlanta, GA 30332-0520; nagesh.murthy@mgt.gatech.edu

Recently, several studies on product development have focused on the interface of marketing, manufacturing and engineering design. However, there is limited work that has considered integrating product development decisions with supply chain management issues. We develop a model for guiding product platform development to support a global value-chain.

SC15.4 Efficient Joint R&D Efforts by Having the Option to Switch Contractual Terms

• Eran Liron; Stanford University, Grad. School of Bus., Stanford, CA 94305; liron@leland.stanford.edu
• Evan L. Porteus; Stanford University, Grad. Sch. of Bus., Stanford, CA 94305-5015; eporteus@leland.stanford.edu

We examine a 2-party, joint R&D problem; one developing the product and the other the custom process, each acting in its self-interest but with possible future transactions. Yields depend on respective investments. We give conditions under which system optimal performance is achieved by an option to switch the sharing rule.

Credit Customers

SC16.1 Expert Probabilistic Systems & Neural Networks in Credit Scoring Analysis

• Maria Arns Steiner; UFPR, Rua Padre Anchieta, 1231 Ap. 31, Curitiba, PR, 80730-000 , Brazil; tere@mat.ufpr.br
• Celso Carnieri; UFPR, Caixa Postal 19002, Curitiba, PR, 81.531-990, Brazil; carnieri@mat.ufpr.br
• Pedro J. Steiner Neto; CEPPAD-UFPR/FAE, Rua Padre Anchieta, 1231 Ap. 31, Curitiba, PR, 80730-000 , Brazil; pedro.steiner@avalon.sul.com.br

With about 2,500 historical data from a German bank, we studied the probabilistic expert systems and neural networks techniques in order to recognize and foresee which credit customers will be 'good or bad payers'. These techniques were used comparatively, using the shell SPIRIT and MatLab-Neural Networks Toolbox to do the computational tests...

SC16.2 Credit Scoring & Evaluation: A Review of Practices & Theories

• Joel K. Jolayemi; University of Zululand, Private Bag X1001, Kwadlangezwa, KwaZulu-Natal, 3886 , South Africa; jkjolaye@pan.uzulu.ac.za

We present a survey and critical review of the procedures and models developed/applied for credit scoring and evaluation. The presentation is divided into 2 parts: the first reviews the procedures applied in some financial institutions while the second surveys and reviews the techniques and models developed in the literature.

SC16.3 Issues in Modeling Lifetime Value of Customer in Credit Card Firms

• Suresh K. Nair; University of Connecticut, Sch. of Bus. Admin., Storrs, CT 06269-0241; suresh@sba.uconn.edu

Modeling is increasingly being used in financial services for risk, marketing and back room operations. We discuss the issues involved in modeling the lifetime value of customer to credit card companies using a Markov decision process approach. Such models may be used for credit line management and other proactive strategies for influencing customer behavior...

Methodological Issues of Planning

SC17.1 Planning & Control with Multi-Resolutional Search, Decomposition & Filtering

• Alex Meystel; Drexel University, Dept. of ECE, Philadelphia, PA 19104; alex@cerebrum.impaqt.drexel.edu

We focus on the inter-relatedness of the stages of planning and control in the functioning of complex systems depending on dealing with large amounts of potentially incomplete and frequently uncertain information. Operators of search, decomposition and filtering function as a package of mutually supportive goal-oriented computational algorithms called GFS-operator...

SC17.2 Linguistic Geometry for Solving Air Combat Games

• Boris Stilman; University of Colorado, Dept. of Comp. Sci. & Eng., CB 109, Denver, CO 80217-3364; bstilman@carbon.cudenver.edu

This method of planning has evolved from the research in computer chess. I was involved in a joint research with the well-known chess-player and engineer M. Botvinnik. Two fundamental steps were undertaken: creation of a planning language that fits within the law of space tessellation and discovery of game horizons that fit into the concept of multi-resolutional control.

SC17.3 Multi-Site Scheduling using Soft Computing Techniques

• Marcello Frixione; DIST - University of Genova, via Opera Pia 13, Genova, I-16145 , Italy; frix@dist.unige.it
• G. Vercelli; University of Trieste, , , Italy; vercelli@sdf.unige.it
• Renato Zaccaria; University of Genova, DIST, via Opera Pia 13, Genova, I-16145 , Italy; renato@dist.unige.it

Multi-site scheduling deals with the scheduling problems of an enterprise with several distributed production sites, where sites use the intermediate products of the other sites to manufacture the products of the enterprise, e.g., in car or airplane manufacturing. In the multi-site scheduling scenario we differentiate between a global and a local scheduling level...

SC17.4 An Open Architecture for Flexible Scheduling Productio & Revision

• Amadeo Cesta; Italian National Research Council, Viale Marx 15, Rome, I-00137 , Italy; amadeo@pscs1.irmkant.rm.cnr.it
• Angelo Oddi; Italian National Research Council, Viale Marx 15, Rome, I-00137 , Italy;

SC17.5 Propagating Search Envelopes in Multi-Resolutional Behavior Planning

• Alex Meystel; Drexel University, Dept. of ECE, Philadelphia, PA 19104; alex@cerebrum.impaqt.drexel.edu
• J. Albus; NIST, Gaithersburg, MD 22899; albus@cme.nist.gov
• A. Lacaze; NIST, Gaithersburg, MD 22899;
• C. Tasoluk; NIST, Gaithersburg, MD 22899;

Innovations & Applications of Integer Programming

SC18.1 An Integrated Production & Transportation Scheduling Problem on a Two-Echelon Manufacturing Network

• Renato de Matta; University of Iowa, 108 Pappajohn Bus. Admin. Bldg, Iowa City, IA 52242-1000; renato-dematta@uiowa.edu
• Tan Miller; Warner-Lambert Inc., 201 Tabor Rd., Morris Plains, NJ 07950; mille15@mops.wl.com

We find the most efficient use of production facilities of a supplier and a manufacturing plant by coordinating their production and shipping schedules. We solve a MIP model that minimizes the total of production, inventory, setup and transportation costs subject to demand and capacity constraints at both facilities.

SC18.2 Short-Term Scheduling in Batch Processing

• Monique Guignard-Spielberg; University of Pennsylvania, The Wharton Sch., 1320 SH-DH, Philadelphia, PA 19104-6366; guignard@grace.wharton.upenn.edu
• Siqun Wang; University of Pennsylvania, The Wharton Sch., 1343 SH-DH, Philadelphia, PA 19104-6366; siqun@assets.wharton.upenn.edu

Production of multiple products on multiple machines is done in batches. We consider identical as well as different batch lengths. We describe a representation similar to STN and algorithmic tools used to improve the solution process. We present computational results on data sets similar to those of Westenberger & Kallrath.

SC18.3 Just-in-Time Scheduling with a Common Due Window

• Zhi-Long Chen; University of Pennsylvania, Dept. of Systems Eng., 220 S 33rd St., Philadelphia, PA 19104-6315; zlchen@seas.upenn.edu
• Chung-Yee Lee; Texas A&M University, 238 Zachry Bldg., Dept. of IE, College Station, TX 77843-3131; cylee@acs.tamu.edu

We propose a column generation based B&B algorithm for the parallel machine JT scheduling problem where all jobs have a common due date and earliness and tardiness penalty weights are job dependent. The algorithm can solve medium size problems within a reasonable CPU time.

SC18.4 Developing Driver Work Schedules for an Inter-City Bus Service Company with Multiple Fleets

• Emmanuel Peters; University of Iowa, Dept of Mgmt Sci, Coll of Bus, 108 Pappajohn BAB, Iowa City, IA 52242; emmanuel-peters@uiowa.edu
• Renato de Matta; University of Iowa, 108 Pappajohn Bus. Admin. Bldg, Iowa City, IA 52242-1000; renato-dematta@uiowa.edu

A company that operates multiple fleets and employs drivers with different skills must decide the mix of fleets to be driven in the driver work schedules. We model this decision as an integer programming problem and propose a column generation approach to solve the problem. Computational results are presented.

Constraint Programming Languages

SC19.1 withdrawn - author request of 10/5

• Yves Caseau; Bouygues, 1 ave. Eugene Freyssinet, St. Quentin Yvelines, 78061 , France; ycs@challenger.bouygues.fr
• Francois Laburthe; Bouygues, 1 ave. Eugene Freyssinet, St. Quentin Yvelines, 78061 , France;
• Glenn Silverstein; Bellcore, 445 South St., Morristown, NJ 09760;

SC19.2 Localizer: A Modeling Language for Local Search

• Laurent Michel; ILOG, 12 rue Charles Bouvier, Bouge, B5004 , Belgium; ldm@ilog.fr

Local search is a traditional technique to solve combinatorial search problems and has raised much interest in recent years. The design and implementation of local search algorithms is not an easy task in general and may require considerable experimentation and programming effort. However, little support is available to assist the design and implementation of local search algorithms...

SC19.3 OPL Script: Composing & Controlling Models

• Pascal Van Hentenryck; Universite Catholique de Louvain, Dept. of Comp. Science & Eng., 2 Place Sainte-Barbe, Louvain la Neuve, B-1348 , Belgium; pvh@info.ucl.ac.be

We give a brief introduction to OPL Script, a script language for composing and controlling optimization models. OPL Script makes it possible to concisely state many applications that require solving several instances of the same model, e.g., to perform some sensitivity analysis, a sequence of models or a combination of both as in column-generation applications...

Traffic Detection & Management Algorithms

SC20.1 An Integer Formulation for SO-DTA to Derive Optimal Control Strategies for Incident Traffic Management

• Omar Sawaya; Northwestern University, Dept. of Civil Eng., 2145 Sheridan Rd., Evanston, IL 60208; o-sawaya@nwu.edu
• Athanasis Ziliaskopoulos; Northwestern University, 2145 Sheridan Rd., Evanston, IL 60208; a-z@nwu.edu

We present an approach that computes optimal control strategies in the form of alternate routes around freeway incident and in response to the prevailing traffic conditions. An integer programming formulation for SO-DTA is introduced, that accounts for the discrete nature of location specific information provided by a VMS.

SC20.2 Optimal Content of Messages on Variable Message Signs for Incident Management

• Srinivas Peeta; Purdue University, Sch. of Civil Eng., West Lafayette, IN 47907;
• Shyam Gedela; Purdue University, Sch. of Civ. Eng., West Lafayette, IN 47907-1284;

Heuristic techniques are analyzed to determine the optimal time-dependent messages to be displayed on Variable Message Signs on-line. A case study involving an ATMS on the Borman Expressway is discussed.

SC20.3 Automatic Incident Detection Algorithms: Calibration & Evaluation

• Josh Peterman; University of Texas, Dept. of Civ. Eng., ECJ 6.2, Austin, TX 78712;
• Hani S. Mahmassani; University of Texas, Dept. of Civil Eng., ECJ 6.2, Austin, TX 78712-1076;
• Carl Haas; University of Texas, Dept. of Civ. Eng., ECJ 6.2, Austin, TX 78712;

The objective of this research is to develop and apply a formal calibration procedure to several incident detection algorithms to determine trade-offs between incident detection ability and false alarm rate. The algorithms are calibrated and tested using traffic data and incident reports generated by the TransGuide Traffic Management System in San Antonio, Texas.

SC20.4 Detection of Traffic & Section-Based Measures

• Carlos Sun; Rowan University, Glassboro, NJ 08028;
• Stephen G. Ritchie; University of California, Dept. of Civil & Env. Eng., Irvine, CA 92697;
• R. Jayakrishnan; University of California, Dept. of Civil & Env. Eng., Irvine, CA 92697;

We present intelligent surveillence technology based on detection of vehicle signatures from point to point in traffic flow, and associated algorithms based on lexicographic pattern matching concepts. We describe the benefits in measuring section variables such as density and travel times, and provide results from a field study.

Path Algorithms & Routing

SC21.1 Properties of Online Shortest Path Problems

• S. Travis Waller; Northwestern University, Dept. of Civ. Eng., 2145 Sheridan Rd., Tech. Inst., Evanston, IL 60208; travis@trans.civil.nwu.edu
• Athanasis Ziliaskopoulos; Northwestern University, 2145 Sheridan Rd., Evanston, IL 60208; a-z@nwu.edu
• Julie Sara Mobley; Northwestern University, Dept. of ISE, 2145 Sheridan Rd., Tech. Inst., Evanston, IL 60208; jsmob@nwu.edu

We discuss shortest path problems where the travel times on links are random variables and vehicles transversing the network acquire information as they progress. We present system properties and specific examples with solution algorithms. Computational results are presented and discussed.

SC21.2 A Multi-Objective Shortest Path Algorithm for Large-Scale Intermodal Transportation Networks

• Khaled F. Abdelghany; University of Texas, Dept. of Civil Eng., ECJ 6.2, Austin, TX 78712;
• Hani S. Mahmassani; University of Texas, Dept. of Civil Eng., ECJ 6.2, Austin, TX 78712-1076;

A multi-objective shortest path algorithm for large scale intermodal transportation networks is presented. The algorithm searches all modes' networks to determine Pareto optimal paths from every node in the network to a single destination. Computational results for 2 different implementations are presented.

SC21.3 Snowplow Districting & Routing

• Xu Jun Eberlein; Caliper Corporation, 1172 Beacon St., Newton, MA 02161;

We present effective heuristic algorithms for designing snowplow districts and routes. Given locations and capacities of plow truck stations, the objectives are to design districts and routes with balanced worktime and minimimal deadhead time. Because a route can be plowed either by a single truck or tandem trucks, the algorithm also determines the best truck assignment to each route.

SC21.4 A Real-Time Routing Algorithm for use in Stochastic, Time-Varying Networks

• Elise D. Miller-Hooks; Pennsylvania State University, 212 Sackett Bldg., Dept. of Civil & Environ. Eng., University Park, PA 16802;
• Baiyu Yang; Pennsylvania State University, Dept. of Civ. & Environ. Eng., 212 Sackett Bldg., University Park, PA 16802; bxy115@psu.edu

One of the primary goals of intelligent transportation systems is to provide travelers with en-route routing suggestions in response to real-time information. An efficient real-time routing algorithm is presented for determining en-route least expected time paths given current and predicted travel conditions, where inherent dynamic and probabilistic characteristics of future conditions are considered.

Negotiating a Y2K Outsourcing Agreement Case

SC22.1 Case 1: Vendor Perspectives

• Michael Jacobides; University of Pennsylvania, The Wharton Sch., Mgmt. Dept., 2055 SH-DH, 3620 Locust Walk, Philadelphia, PA 19104;

No abstract supplied.

SC22.2 Case Two: Client Perspectives

• David C. Croson; University of Pennsylvania, The Wharton School, Philadelphia, PA 19104;

No abstract supplied.

Longer-Run Issues in Electricity Restructuring

SC23.1 LCP Models of Nash-Cournot Competition in the England & Wales Electricity Market

• Christopher J. Day; London Business School, Sussex Place, Regents Park, London, NW1 4SA , England, UK; cday@lbs.ac.uk
• Benjamin F. Hobbs; JHU, 313 Ames Hall, DOGEE, 3400 North Charles St., Baltimore, MD 21218; bhobbs@jhu.edu

An LCP formulation of competition between producers of electricity in England and Wales is considered. We investigate issues relating to the reform of this market, from its present poolco mechanism to bi-lateral trading.

SC23.2 Ensuring Generation Capacity Adequacy in Competitive Electricity Markets

• Shmuel S. Oren; University of California, Dept. of IE/OR, 4172 Etcheverry Hall, Berkeley, CA 94720-1777; oren@ieor.berkeley.edu

I will discuss alternative approaches that have been adopted around the world for guaranteeing the appropriate level of investment in electric generation capacity. I argue that 'capacity payment' is the least desirable approach and it undermines the long-term efficiency objectives of the electric industry restructuring.

SC23.3 A Generation Planning Model in Electricity Markets Based on the Complementary Problem

• Michel Rivier; Universidad Pontificia Comillas, Inst. de Investigacion Tec., Alberto Aguilera 23, Madrid, 28015 , Spain; michel.rivier@iit.upco.es
• Mariano Ventosa; Universidad Pontificia Comillas, Inst. de Investigacion Tec., Alberto Aguilera 23, Madrid, 28015 , Spain; mventosa@iit.upco.es
• Andres Ramos; Universidad Pontificia Comillas, Inst. de Investigacion Tec., Alberto Aguilera 23, Madrid, 28015 , Spain;

The electricity market behavior will depend on companies' production cost profiles, on companies' technical operation constraints, and on the market equilibria resulting from the interaction of all participants. This presentation proposes a new methodology for long term operation planning models, fully adapted to represent an annual or hyperannual generation scheduling in a competitive framework...

SC23.4 The Long-Term Effect of Imperfect Competition in Electricity Generation

• Frederic H. Murphy; Temple University, 2402 Pine St., School of Bus. & Mgmt., Philadelphia, PA 19122; v5256e@vm.temple.edu
• Yves Smeers; Universite Catholique de Louvain, CORE, 34 Voie du Roman Pays, Louvain la Neuve, 1348 , Belgium; smeers@core.ucl.ac.be

We discuss the consequences of capacity expansion in oligopolistic electricity markets. We look at different games, discuss their adequacy for capturing interactions in the market place and derive properties of their solution.

Management Issues in Health Care Technology Management II

SC24.1 A Six-Dimensional Meta-Analysis of the Management of Health Care Technology Literature (1979-97)

• Ashok Kumar; Grand Valley State University, 510 Eberhard Ctr., 301 West Fulton, Grand Rapids, MI 49504; kumara@gvsu.edu
• Jaideep G. Motwani; Grand Valley State University, 510 Ebergard Ctr., 301 West Fulton, Grand Rapids, MI 49504; motwanij@gvsu.edu
• Arnold Reisman; Sabanci University, Grad. Sch. of Mgmt., Istanbul, , Turkey; 102126-1551@compuserve.com

We take a retrospective look at trends, patterns and directions taken by researchers in Management of Health Care Technology during the period 1979-97. Three hundred and thirty three publications are each classifed in 6 distinct categories.

SC24.2 Embeddedness Matrix, the Notion of Technology-Based Assets & a Test of Alternative Theories of Boundary Choices in Medical Technologies

• Eliezer Geisler; Illinois Institute of Technology, Stuart Grad. School of Bus., 565 West Adams St., Chicago, IL 60661; geisler@stuart.iit.edu

Health care organizations have a choice in deciding on acquisition of assets in which medical technology is embedded. We report on preliminary results from a study that tested the transaction cost, knowledge-based and resources-based theories of the firm to explain decisions by health care organizations in their consideration of medical technology...

Queueing Networks

SC26.1 Strong Approximations for a Feedforward Multiclass Queueing Network

• Xinyang Shen; University of British Columbia, Fac. of Commerce & Bus. Admin., 2053 Main Mall, Vancouver, BC, V6T 1Z2 , Canada; xshen@hong.commerce.ubc.ca
• Hong Chen; HKUST, Dept. of IE/EM, Clear Water Bay, Kowloon, , Hong Kong; hongchen@uxmail.ust.hk

We obtain a strong approximation for a feedforward multiclass queueing network that provides a refinement for the diffusion approximation and doesn't require a heavy traffic condition. We propose some specific approximations that lead to numeric approximations to the workload, the queue length and the sojourn time processes of the queueing network.

SC26.2 Scheduling Multiclass Networks in Heavy Traffic: Asymptotic Optimality via Discrete-Review Policies

• Costis Maglaras; Columbia University, Columbia Business Sch., 409 Uris Hall, 3022 Broadway, New York, NY 10027-6902; c.maglaras@columbia.edu

For a multiclass network under heavy loading, an approximating Brownian control problem is formulated and solved. An implementable discrete-review policy is defined based on the optimal Brownian solution. Using the Bramson-Williams framework, we establish asymptotic optimality in the heavy-traffic sense by analyzing the fluid behavior under this policy.

SC26.3 Queueing Networks with State Dependent Transitions

• Masakiyo Miyazawa; Science University of Tokyo, Dept. of Info. Sciences, Noda-City Chiba, 278 , Japan; miyazawa@is.noda.sut.ac.jp
• Xiuli Chao; NJIT, Dept. of IE, University Heights, Newark, NJ 07102; chao@admin.njit.edu,, chao@megahertz.njit.edu

We give a procedure to construct a queueing network with network-state dependent local transitions from a quasi-reversible queueing network state independent local transitions. We show that under some conditions of additional departures, the state-dependent queueing network has a tractable stationary distribution. This stationary distribution, under some conditions, is a stochastic upper bound of the original network.

Perturbation Analysis & Simulation II

SC27.1 A Perturbation Analysis-Based Solution to the Ground-Holding Problem in Air Traffic Control

• Christos G. Panayiotou; University of Massachusetts, Dept. of Elect. & Comp. Eng., Amherst, MA 01003; panayiot@ecs.umass.edu
• Christos G. Cassandras; Boston University, Dept. of Mfg. Eng., 15 Saint Mary's St., Boston, MA 02215; cgc@bu.edu

We propose an efficient, real-time approach for determining the optimal ground-holding delay (GHD) for every aircraft. Using finite perturbation analysis, we predict the effect of adding a new aircraft to an existing schedule of arriving flights and hence determine the GHD that minimizes some cost function.

SC27.2 IPA Derivative Estimation for a Queueing System

• Pirooz Vakili; Boston University, Dept. of Mfg. Eng., 15 Saint Mary's St., Boston, MA 02215; vakili@bu.edu

We propose an algorithm for estimating performance derivatives with respect to routing parameters in a queueing system and discuss the statistical properties of the estimators. We give a review of other algorithms that have been used to solve the same problem.

SC27.3 A Global Approach for Large-Scale Buffer Allocation Problems

• Leyuan Shi; University of Wisconsin, Dept. of IE, Madison, WI 53706; leyuan@ie.engr.wisc.edu

We present a hybrid approach that combines perturbation analysis techniques with the nested partition method for a large scale buffer allocation problem. The effectiveness of the hybrid algorithm will be demonstrated through a numerical example.

Rigorous Qualitative Analysis

SC28.1 Y2K: Death of Simulations Visions of the Future

• Gerald Jones; California State University, Dept. of IS/DS, Craig Sch. of Bus., Fresno, CA 93740-8001; geraldj@csufresno.edu

While much has been heard about the problems created by the Y2K bug, most academics expect it to have little, if any, impact on their normal activities. Unfortunately, that is not always the case. We report on the impact the bug will have on batch-processed simulations and alternatives to be developed.

SC28.2 Tutorial/Workshop in Leadership

• Gerald Jones; California State University, Dept. of IS/DS, Craig Sch. of Bus., Fresno, CA 93740-8001; geraldj@csufresno.edu

This workshop allows participants to experience the use of a management ethics survey in exploring differences in responses to ethical situations as a function of generation and various other demographic variables. Participants will find the material and process employed useful in leadership, ethics, management research and in the classroom for generating discussion and learning.

Public Policy Applications

SC29.1 Crime Displacement Phenomenon & Enforcement Allocation Policies

• Shoou-Jiun Wang; SUNY, 342 Bell Hall, Dept. of IE, Buffalo, NY 14260; sw@eng.buffalo.edu
• Rajan Batta; SUNY, 342 Bell Hall, Box 602050, Dept. of IE, Buffalo, NY 14260; batta@acsu.buffalo.edu
• Christopher M. Rump; SUNY, 342 Bell Hall, Dept. of IE, Buffalo, NY 14260-2050; crump@eng.buffalo.edu

Criminal displacement is the phenomenon when offenders of certain types of crimes shift the incidence of crime to other types of crimes, times or locations. We develop rules of thumb aimed at effective policy design through an understanding of how criminal displacement occurs and improving enforcement operations.

SC29.2 Workflow Modeling for Inter-Organizational Virtual Information Dissemination

• Siva R. Sankaran; California State University, Dept. of Acct. & MIS, 18111 Nordhoff St., Northridge, CA 91330-8372; siva@pacbell.net
• Tung X. Bui; University of Hawaii, 2404A Maile Way, E303A, Honolulu, HI 96822; tbui@busadm.cba.hawaii

We propose the concept of virtual information dissemination and advocate a domain-independent architecture for a system that, when implemented on the Internet, can serve as a decision support tool for those managing a humanitarian crisis. Expectations in terms of accuracy, response time and reduction of human suffering were met.

SC29.3 Blacklisting Social Science Institutions: A Revisit

• Richard A. Colombo; Fordham University, 113 West 60th St., New York, NY 10023; rcolombo@pipeline.com
• Don G. Morrison; UCLA, 405 Hilgard Ave., Los Angeles, CA 90024; dmorrison@anderson.ucla.edu
• Robin Wensley; Warwick University, Gilbet Hill Road, Warwick, CV4 7AL , UK; msmjrw@raxor.wbs.warwick.ac.uk

We examine the effect of the ESRC's policy of denying PhD studentships to institutions whose submission rates fall below a threshold. We use a beta-binomial model with time-varying parameters to investigate how submission rates from over 80 UK social science institutions have changed over the past 10 years.

SC29.4 Implementing Cost-Effective Technology Training Programs at Public Schools

• Yun Wang; Mercy College, Math/CIS Dept., 555 Broadway, Dobbs Ferry, NY 10522; ywang@mercynet.edu

A better technological environment for this nation's public education relies on many factors. Technology training is one factor, playing a critical role in improving district's education standards in our current high-tech world. We provide various ways to implement cost-effective technology training processes within a district's existing tight budget.

Aviation System Safety & Risk Analysis

SC31.1 System Safety & Risk Analysis for Air Carrier Oversight

• Donald Arendt; Federal Aviation Administration, Flight Standards Certification, AFS 900, 45005 Aviation Dr., Dallas, VA 20166; don.arendt@faa.dot.gov

We focus on the rule of system safety principles and risk analysis techniques in the development of surveillance and oversight programs for aircraft carriers.

SC31.2 withdrawn - author request of 10/11

• Kathleen V. Diegert; Sandia National Lab., Stats. & Human Factors Dept., PO Box 5800, Albuquerque, NM 87185-0829; kvdiege@sandia.gov
• Karen S. Marlman; Sandia National Lab., Stats. & Human Factors Dept., PO Box 5800, Albuquerque, NM 87185-0829; ksmarlm@sandia.gov
• Alice J. Johnson; Sandia National Lab., Stats. & Human Factors Dept., PO Box 5800, Albuquerque, NM 87185-0829; ajjohns@sandia.gov

SC31.3 Monitoring Multiple Aviation Safety Data by Data Depth: Control Charts & Threshold Systems

• Regina Liu; Rutgers University, Dept. of Statistics, Hill 453, Piscataway, NJ 08854; liu@stat.rutgers.edu
• James Luxhoj; Rutgers University, Dept. of IE, 96 Frelinghuysen Rd., Piscataway, NJ 08854; luxhoj@princess.rutgers.edu
• Andrew Cheng; Federal Aviation Administration, FAA Technical Ctr., Tilton Rd., Bldg. 300, Atlantic City Airpor, NJ 08405; andrew_ctr_cheng@faa.gov

We present an application of a nonparametric methodology based on data depth for monitoring multivariate safety observations. The control chart is combined here with threshold levels for evaluating the performance of aircraft operators and provides clear guidelines for identifying unexpected performances and for aligning appropriate corrective actions.

SC31.4 Applying Hilbert-Space Methods to Multivariate Quality Control & Risk Monitoring

• Lars Nordmann; Rutgers University, Dept. of IE, 96 Frelinghuysen Rd., Piscataway, NJ 08854;
• James Luxhoj; Rutgers University, Dept. of IE, 96 Frelinghuysen Rd., Piscataway, NJ 08854; luxhoj@princess.rutgers.edu

We explore the utility of Hilbert-Space methods for the statistical analysis of multivariate data clouds. Applications can be found in aviation safety and general risk assessment.

High-Speed Networks

SC32.1 Routing in ATM Networks

• Pascal R. Chanas; France Telecom CNET, 905 rue Albert Einstein, Sophia-Antipolis, 06921 , France;
• Olivier Goldschmidt; Make Systems, 4000 Regnecy Park, Cary, NC 27511; olivier@makesys.com
• Michel Burlet; Laboratoire LEIBNIZ-IMAG, 46 av. Felix Viallet, Grenoble Cedex, 38031 , France;

We consider the arc-disjoint path problem where demands in an ATM network must be routed on a unique path. We present new theoretical results for the problem in symmetric digraphs and heuristics to solve routing problem for VP network design. We present empirical results from real world cases.

SC32.2 Virtual Path Layout in ATM Networks

• Olivier Goldschmidt; Make Systems, 4000 Regnecy Park, Cary, NC 27511; olivier@makesys.com
• Michel Burlet; Laboratoire LEIBNIZ-IMAG, 46 av. Felix Viallet, Grenoble Cedex, 38031 , France;
• Pascal R. Chanas; France Telecom CNET, 905 rue Albert Einstein, Sophia-Antipolis, 06921 , France;

We present a mathematical model and algorithmic techniques for optimizing the layout of virtual paths in an ATM network with the general objective of minimizing the number of VPs between nodes in the network. Specifically, we design a virtual path network that minimizes the diameter of the VP network.

SC32.3 The Path Restoration Version of the Spare Capacity Allocation Problem with Modularity Restrictions

• Jeffery L. Kennington; SMU, Dept. of Comp. Sci. & Eng., Dallas, TX 75275-0122; jlk@seas.smu.edu
• Mark W. Lewis; SMU, Dept. of Comp. Sci. & Eng., Dallas, TX 752114; mlewis@seas.smu.edu

We present an efficient algorithm for the path restoration version of the spare capacity allocation problem with modularity restrictions. The algorithm is a special implementation of B&B that exploits modularity. Special cuts are used to obtain a tighter LP relaxation. An empirical analysis is presented.

SC32.4 The SONET Edge Partition Problem

• Eli V. Olinick; University of California, Department of IEOR, 4135 Etcheverry Hall, Berkeley, CA 94720-1777; olinick@ieor.berkeley.edu, http://www.ieor.berkeley.edu/~olinick
• Olivier Goldschmidt; Make Systems, 4000 Regnecy Park, Cary, NC 27511; olivier@makesys.com
• Dorit S. Hochbaum; University of California, Haas Sch. of Bus., Dept. of IEOR, 4135 Etchverry, Berkeley, CA 94720-1777; dorit@hochbaum.berkeley.edu, http://www.ieor.berkeley.edu/~hochbaum/

Given pairwise bandwidth demands, we wish to connect customer sites with a network of unidirectional SONET rings. Sites may be assigned to multiple rings; however, traffic between 2 sites cannot be split between rings. The objective is to minimize the total number of ADMs required for the network.

Predictive & Proactive Maintenance: Models & Methods

SC33.1 Predictive Maintenance using a Precursor Window

• Lisa M. Maillart; University of Michigan, Dept. of IOE, 1205 Beal Ave., Ann Arbor, MI 48109-2117; maillart@umich.edu
• Stephen M. Pollock; University of Michigan, Dept. of IOE, 1205 Beal Ave., Ann Arbor, MI 48109-2117; pollock@umich.edu

If condition monitoring of a machine occurs within a precursor window, preventive maintenance can be performed that time. However, if failure occurs before monitoring, reactive maintenance must be performed. We find optimal monitoring intervals that balance between costs of monitoring and maintenance. The results suggest a practical method for allocating monitoring resources among many machines.

SC33.2 PM-Expert: A Decision Support System for Profit-Driven Maintenance

• Daniel J. Reaume; General Motors R&D Center, 30500 Mound Rd., Warren, MI 48090-9055; daniel_reaume@gmrnotes3.gmr.com
• Jeffrey M. Alden; General Motors R&D Center, 30500 Mound Rd., Warren, MI 48090-9055; jeff_alden@gmrnotes3.gmr.com

Balancing increased profit against the cost of maintenance is a critical problem for managers of production systems. A DSS is needed to analyze the effects of maintenance planning decisions on safety, quality, throughput and profit. We present such an application, PM-Expert, for the user-friendly preparation, analysis and graphical presentation of maintenance related business cases.

SC33.3 Coordinating Quality Control & Maintenance

• Kevin W. Linderman; University of Minnesota, 3-235 Carlson Sch. of Mgmt., Minneapolis, MN 55455; klinderman@csom.umn.edu
• Kathleen E. McKone; University of Minnesota, 3-235 Carlson Sch. of Mgmt., Minneapolis, MN; kmckone@csom.umn.edu

Process variation is tightly linked to the condition of the production equipment. Therefore, control of the process may be improved if quality control methods are supplemented with equipment maitenance policies. In particular, we recommend that maintenance costs...

The Theory & Applications of Supply Chain Management

SC34.1 Selection of Suppliers Considering the Learning Effect & Technology Improvement

• Shanling Li; McGill University, Faculty of Management, 1001 Sherbrooke St. West, Montreal, Quebec, H3A 1G5 , Canada; li@management.mcgill.ca
• Wendy Huang; Lakehead University, Dept. of Math & Stats., Sch. of Math. Science, Thunder Bay, Ontario, P7B 5E1 , Canada;
• Devanath Tirupati; University of Texas, Dept. of Mgmt., CBA 4.202, Austin, TX 78712;

We examine the impact of supplier learning and technology improvement and compare long-term and short-term suppliers. Signing a long-term contract will be benefited from learning effect. Periodically searching the best market price may take the advantage of technology breakthroughs. We developed models to describe the supplier selection problems and derive analytical results.

SC34.2 System Cooperation with Quantity Discount Consideration

• Zhimin Huang; Adelphi University, Sch. of Mgmt. & Business, 1 South Ave., Box 701, Garden City, NY 11530; huang@panther.adelphi.edu
• Susan Li; Adelphi University, School of Business, 1 South Ave., Box 701, Garden City, NY 11530; susan@panther.adelphi.edu

This study deals with a situation where the buyer has a monopolistic position, and examines the issues and advantages of cooperation in a buyer-seller inventory control system. Game theory concepts form the foundation for our analysis of these issues. The similarity and differences between our model and those in the literature are discussed.

SC34.3 Approaches to a Mail Transportation Network Problem

• Wen-Chyuan Chiang; University of Tulsa, Finance & OM Dept., 600 South College Ave., Tulsa, OK 74104-3189; wen-chyuan-chiang@utulsa.edu
• Panagiotis Kouvelis; Washington University, Olin Sch. of Business, 1 Brookings Dr., St. Louis, MO 63130-4899;

We consider a mail network problem with different transportation modes. We propose intelligent heuristics to solve the problem. The quality of computational results is demonstrated.

Tutorial: Generalized Hill Climbing Algorithms in Optimization

SC35.1 Tutorial: Generalized Hill Climbing Algorithms in Optimization

• Sheldon H. Jacobson; University of Illinois, Dept. of MIE, 1206 West Green St., MEB 226, Urbana, IL 61801-2906; shj@uiuc.edu,, http://www.staff.uiuc.edu/~shj/shj.html

GHC algorithms have been introduced as a unifying framework for addressing intractable discrete optimization problems. GHC algorithms provide a well-defined structure for classifying and studying a large body of stochastic and deterministic search strategies. Simulated annealing, threshold accepting and TS, among others, can all be formulated as particular GHC algorithms. We review the GHC structure...

DEA II

SC36.1 Determining Academic Departments Productivity & Quality with a Fuzzy Data Envelopment Analysis Model

• Ana Lucia Lopes; Universidade do Sul de Santa Catarina, Rua do Guapuruvu, 155, Florianopolis, SC, 88062-240 , Brazil; ana@eps.ufsc.br
• Edgar Augusto Lanzer; Universidade do Sul de Santa Catarina, Av. Jose Acacio Moreira, 787, Tubarao, SC, 88701-900 , Brazil; ealanzer@unisul.rct-sc.br
• Ricardo M. Barcia; Universidade Federal de Santa Catarina, UFSC/EPS, Campus Universitario Trindade, Florianopolis, SC, 88040-900 , Brazil; barcia@eps.ufsc.br

A model to evaluate the performance productivity and quality of academic departments is proposed, based on DEA that was applied to sets of teaching, research, service and quality indicators resulting in 4 frequency distributions that can be interpreted as fuzzy productivities measures. The fuzzy measures were integrated through an optimum (simulated) department choice of an ordered weighted aggregator...

SC36.2 Estimation of Efficiency Score under Uncertainty

• Hideo Michida; Kobe University, Dept. of CSE, Rokkodai-cho, Nada, Kobe, Hyogo, 657-8501 , Japan; michida@kaede.seg.kobe-u.ac.jp
• Hiroshi Morita; Kobe University, Dept. of Comp. & Systems. Eng., Rokkodai-cho, Nada, Kobe, Hyogo, 657-8501 , Japan; morita@seg.kobe-u.ac.jp
• Noriyoshi Nakayama; University of Marketing & Distrib. Science, Gakuen-nishimachi, Nishi, Kobe, Hyogo, 6512188 , Japan; noriyoshi_nakayama@red.umds.ac.jp

The efficiency score calculated using uncertain data is also uncertain, and the inefficiency should be included the inefficiency from uncertainty. We propose an interval of efficiency score to estimate the true efficiency score which is not spoiled by uncertainty. The uncertain data are given as repeated observations in this study.

SC36.3 An Interactive Approach to Improve Estimates of Value Efficiency in Data Envelopment Analysis

• Tarja Joro; University of Alberta, 3-04F Business Bldg., Edmonton, Alberta, T6G 2R6 , Canada; tjoro@ualberta.ca,, http://www.bus.ualberta.ca/tjoro
• Pekka J. Korhonen; Helsinki School of Economics, PO Box 1210, Helsinki, 00101 , Finland; korhonen@hkkk.fi,, http://www.hkkk.fi/~korhonen
• Stanley Zionts; SUNY, School of Mgmt., Jacobs Mgmt. Ctr., Buffalo, NY 14260; szionts@acsu.buffalo.edu

Value efficiency analysis is an approach to incorporate preference information in DEA. Some related concepts to value efficiency analysis are developed in order to obtain more precise estimates. DEA, value efficiency analysis and the proposed refinements are then applied to estimate the efficiency of university research units.

SC36.4 Non-Parametric Analysis of Productive Efficiency using Engineering Knowledge & Detailed Data

• David W. Moore; University of California, 4135 Etcheverry Hall, Dept. of IEOR, Berkeley, CA 94720; davidm@ieor.berkeley.edu

We develop a non-parametric theory of productive efficiency for detailed engineering data, as opposed to aggregate economic data. We endow the input and output correspondences with additional structure compared to those of traditional DEA models. This additional structure leads to new modes of efficiency analysis and generalizes the DEA approach.

Integrating OR/MS & Finance in the Classroom

SC37.1 Teaching Finance-Related Decision Modeling

• Samuel E. Bodily; University of Virginia, Darden Grad. Sch. of Bus. Adm., Box 6550, Charlottesville, VA 22906-6550; bodilys@darden.gbus.virginia.edu

Teaching finance-related decision modeling in the MBA quantitative required course, in MBA electives, and in executive education is discussed. Topics include appraisal of capital projects in uncertainty, risk management and preference, how to create value and reduce risk, staged investing, and hedging. The approach uses cases and spreadsheets with simulation and other add-ins.

SC37.2 A Program in Investment Science

• David G. Luenberger; Stanford University, EES&OR Dept., 314 Terman Eng. Ctr., Stanford, CA 94305-4023; luen@leland.stanford.edu

I shall describe an integrated educational program consisting of several courses beginning at the MS level and progressing through the PhD. The series begins with the three most important analytic techniques and builds a solid foundation of theory and computation methods, applied to both securities and real projects.

SC37.3 Integrating Finance & MS in an MBA Program

• Wayne L. Winston; Indiana University, Kelley School of Bus., Bloomington, IN 47405; winston@indiana.edu

We will discuss how finance and management science are integrated in the IU MBA program. This fall I am giving eight lectures in the second year finance electives which integrate finance and MS. The content of these courses will be discussed.

Utilities for Groups & Corporations

SD01.1 Multiperson Utility

• Manel Baucells; UCLA, The Anderson School, 655 Levering Ave., Los Angeles, CA 90024; mbaucell@anderson.ucla.edu
• Lloyd Shapley; UCLA, Box 951477, Los Angeles, CA 90095-1477; shapley@econ.ucla.edu

We set the problem of preference aggregation in a context of incomplete preferences. The main result is that a hierarchy (formally, a spanning tree) of bilateral agreements along with the extended Pareto rule is sufficient to build a complete preference for groups.

SD01.2 Risk Tolerances for Quasi-Syndicates & Publicly Held Firms

• Peter C. Anselmo; New Mexico Institute of Mining & Technology, Dept. of Mgmt., Speare Hall, Box 3, Socorro, NM 87801; anselmo@nmt.edu
• James S. Dyer; University of Texas, Dept. of MSIS, Coll. of Bus. Admin., Austin, TX 78712; jim.dyer@bus.utexas.edu

We present an extension of the classic risk tolerance aggregation result of Wilson (1968) in the context of a financial market where investors hold many different portfolios. Our market-based analysis is compared with observed and assessed risk tolerances for 15 publicly-traded energy firms.

SD01.3 The Impact of Risk Attitude Approximations on Decision Analysis Results

• Craig W. Kirkwood; Arizona State University, Dept. of Management, Tempe, AZ 85287-4006; craig.kirkwood@asu.edu

There is a 'folk theorem' in decision analysis practice that says risk aversion doesn't matter in many practical decisions, and that when risk attitude does matter it is often adequate to use an exponential utility function. We report the results of an extensive simulation test of this conjecture.

SD01.4 Risk-Sharing & Corporate Risk Attitudes

• James E. Smith; Duke University, Fuqua Sch. of Bus., Box 90120, Durham, NC 27708-0120; jes9@mail.duke.edu

We consider the problem of deriving a corporate utility function from the risk preferences of the company's shareholders. Though we cannot uniquely determine the corporate utility function (except under very special circumstances), we can place bounds on corporate risk tolerances and certainty equivalents based on the preferences of the shareholders.

Dynamic Routing

SD02.1 A New Model for Dynamic Route Guidance

• Andreas S. Schulz; MIT, Sloan Sch. of Management, E53-361, Cambridge, MA 02139;
• Olaf Jahn; Technical University of Berlin;
• Rolf H. Moehring; Technical University of Berlin;

The design of route-guidance algorithms faces a dilemma. The approach that theoretically yields the optimum system performance may discriminate against some users. Proposed alternate methods, however, do not directly address the system optimum and may result in inferior performance. We propose a novel model and algorithms to resolve this dilemma.

SD02.2 Fluid Analysis of Arrival Routing

• Dimitris Bertsimas; MIT, Sloan Sch. of Mgmt., 50 Memorial Dr., Bldg. E53-363, Cambridge, MA 02142; dbertsim@aris.mit.edu
• Michael H. Veatch; Gordon College, Dept. of Math., Wenham, MA 01984;

Arrivals are routed to one of n queues to minimize average holding cost. Analysis using fluid limits shows that, when queues are large, routing only switches to queues with larger holding cost. The optimal fluid policy is compared with the optimal policy for exponential interarrival and service times.

SD02.3 The Time-Dependent Quickest Transshipment Problem

• Sarah Stock Patterson; Duke University, Fuqua Sch. of Bus., Box 90120, Durham, NC 27708-0120; sara.sp@duke.edu
• Elise D. Miller-Hooks; Pennsylvania State University, 212 Sackett Bldg., Dept. of Civil & Environ. Eng., University Park, PA 16802;

We present a pseudopolynomial time algorithm to solve the time-dependent integral quickest transshipment problem. This algorithm determines the minimum time paths along which to send supply between multiple sources and sinks in a time-dependent (w.r.t. travel times, arc and node capacities and supply) and dynamic (w.r.t. flow) network.

Learning & Testing in Product Development

SD03.1 A Dynamic Project Prioritization Policy

• Stylianos Kavadias; INSEAD, Blvd. de Constance, Fountainbleau, 77305 , France; stylianos.kavadias@insead.fr
• Christoph H. Loch; INSEAD, Blvd. de Constance, Fountainbleau, 77305 , France; christoph.loch@insead.fr

Scarce resources present a typical problem in firms pursuing multiple product development projects in parallel. Once projects are under way and compete for resources, how should priorities be dynamically assigned to maximize total project value? We develop a multi-armed bandit formulation of the problem and derive simple prioritization policies that are continuosly updated.

SD03.2 Learning & Process Improvement during Production Ramp-Up

• Christian Terwiesch; University of Pennsylvania, The Wharton Sch., Dept. of OPIM, SH-DH 1300, Philadelphia, PA 19104; terwiesch@wharton.upenn.edu
• Roger Bohn; University of California, IR/PS 0510, 9500 Gilman Dr., La Jolla, CA 92093-0519;

We analyze the interactions among production speed, capacity utilization, yields and process improvement (learning) during production ramp-up of a new product. We model learning in the form of deliberate experimentation. Experiments such as engineering trials lead to process improvements, but they also reduce capacity in the short run.

SD03.3 Optimal Testing in Product Development

• Stefan Thomke; Harvard Business School, Morgan Hall T63, Soldiers Field, Boston, MA 02163; sthomke@hbs.edu
• David Bell; Harvard Business School, Morgan Hall 171, Soldiers Field, Boston, MA 02163;

A fundamental problem in managing product development is the optimal timing, frequency and fidelity of testing activities that are carried out to evaluate novel product concepts and designs. We will present several important findings on optimal testing strategies that are the result of a formal mathematical model on testing.

SD03.4 Preliminary Information, Testing & Communication in Product Development

• Christoph H. Loch; INSEAD, Blvd. de Constance, Fountainbleau, 77305 , France; christoph.loch@insead.fr
• Christian Terwiesch; University of Pennsylvania, The Wharton Sch., Dept. of OPIM, SH-DH 1300, Philadelphia, PA 19104; terwiesch@wharton.upenn.edu
• Stefan Thomke; Harvard Business School, Morgan Hall T63, Soldiers Field, Boston, MA 02163; sthomke@hbs.edu

We model a product design activity as search in a problem tree, where multiple tests are used to identify a design solution. The problem structure determines the best search policy. It is common in concurrent engineering that a dependent 'downstream' activity must make commitments before the design solution is known...

Multi-Unit Auctions

SD04.1 Trading Portfolios: An Experimental Evaluation

• Sayee Srinivasan; OptiMark Technologies Inc., 530 Main Ave., Durango, CO 81301; ssrinivasan@mailer.optimark-tech.com
• Jan Stallaert; University of Texas, Dept. of Info. Systems, CBA 5.202, Austin, TX 78712;
• Andrew B. Whinston; University of Texas, Coll. of Bus. Admin., CBA 5.202, Austin, TX 78712-1175; abw@utexas.edu

Investors prefer to hold diversified portfolios of assets. We take this dictum seriously and propose a double auction model that would allow investors to trade non-standard combinations of assets. Preliminary results from laboratory experiments indicate that the model can be improved along numerous dimensions. The proposed framework is an initial effort to design markets that reflect true needs of investors.

SD04.2 Auctions & Subsidies

• Indranil Chakraborty; Instituto Technologico Autonomo de Mexico, Dept. of Bus. Admin., Rio Hondo 1 Col. Tizapan-Angel, Mexico DF, 01000 , Mexico; indranil@eniac.rhon.itam.mx

We demonstrate the effect of a subsidy and quantity discount in a multi-unit auction under the (Friedman) uniform price rule. When there are 2 units on sale, a sufficiently large lump-sum subsidy to a bidder who wins both units makes bidders more aggresively on the second unit. This increases the expected revenue to the seller...

SD04.3 Demand Reduction in Multi-Unit Auctions: Evidence from a Sportscard Field Experiment

• David Lucking-Reiley; Vanderbilt University, Nashville, TN; reiley@vanderbilt.edu
• John A. List; University of Central Florida;

Auction theory suggests that multi-unit uniform-price auctions entail demand reduction incentives which may cause inefficient allocations and reduced revenue. We present the results of an experiment with multi-unit auctions for $10,000 worth of sportcards. As predicted theoretically, we find significant demand reduction. The bid differences are large enough to affect the allocation of goods...

Knowledge Representation in DA

SD05.1 A Causal Mapping Approach to Constructing Bayesian Networks

• Sucheta S. Nadkarni; University of Kansas, School of Business, Lawrence, KS 66045; snadkarni@bschool.wpo.ukans.edu
• Prakash P. Shenoy; University of Kansas, School of Bus., Summerfield Hall SU318, Lawrence, KS 66045-2003; pshenoy@ukans.edu,, http://stat1.cc.ukans.edu/~pshenoy

We describe a systematic procedure to construct Bayesian networks based on expert knowledge using the causal mapping approach. We outline how causal beliefs of experts in specific domains can be represented as graphical structures and how probability encoding techniques can be used to derive the parameters of the structure.

SD05.2 Sensitivity of Correlations for Expert Probability Distributions

• Christopher J. Lacke; Rowan University, Math Dept., Glassboro, NJ 08028; lacke@rowan.edu,, http://www.rowan.edu/mars/depts/math/lacke/lacke1.html
• Robert T. Clemen; Duke University, Fuqua Sch. of Bus., Box 90120, Durham, NC 27708-0120; clemen@mail.duke.edu,, http://www.duke.edu/~clemen/web.htm

Using a copula to combine expert probability distributions requires assessing dependence among the experts' opinions. In the context of a colorectal cancer screening model, we study the sensitivity of experts' combined distributions to subjectively assessed correlations. We also study the sensitivity of the choice of the optimal screening regimen.

SD05.3 withdrawn - author request of 9/6

• Roberto Ley-Borras; Insttituto Tecnologico de Orizaba, Oriente 13 A No. 1122, Entre Norte 28 y 28 A, Orizaba, Veracruz, 94300 , Mexico;

SD05.4 A Sample Preparation Expert-Database System for Microwave Digestion III, Object-Oriented Programming & Online Database

• Fei Qin; Duquesne University, 353 Gross St., Apt. 2, Pittsburgh, PA 15224; qin6313@duq.edu
• Howard M. Kingston; Duquesne University, 4000 Forbes Ave., Pittsburgh, PA 15282;

We describe the architecture of a hybrid expert-database system for microwave sample preparation: MWDXpert from a programming point of view. Microsoft Access database tables are used for storing both analytical data and inference rules. Microsoft Excel is used for statistical calculation. Structured query language and active data objects are used to achieve database software independence...

Financial Decision Making in Health Care

SD06.1 The Role of Provider Financial Incentives in Health Care

• Mark Pauly; University of Pennsylvania, The Wharton School, Pennsylvania, PA 19104; pauly@wharton.upenn.edu

This talk will provide an overview of our current knowledge about the role of provider financial incentives in health care.

SD06.2 A Budget Allocation Model for Non-Profit Hospitals Based on DEA

• Jens Muller; ;
• Stefan Scholtes; University of Cambridge, Dept. of Engineering, Cambridge, CB2 1PZ , UK; ss248@cam.ac.uk

We consider n DMUs with corresponding output vectors A1,...,An. We wish to allocate a given Budget B to the DMUs, i.e. to find a budget allocation vector b with b1+...+bn=B. We present a fixed point method for the calculation of this allocation and give some preliminary numerical results.

SD06.3 Modeling the Impact of Reimbursement Policy on Clinics, Patients & Pharmaceutical Firms

• Kut C. So; University of California, Grad. School of Mgmt., Irvine, CA; rso@gsm.uci.edu
• Chris Tang; UCLA, Anderson School of Mgmt., 110 Westwood Plaza Box 951481, Los Angeles, CA 90095-1481;

We present a mathematical model that enables us to examine the impact of a drug reimbursement policy. Despite the simplistic structure of our model, the analysis helps our understanding of the joint impact of such a policy on the patients, the clinic and pharamaceutical firm, and thus can provide valuable information for evaluating the effectiveness of implementing such a reimbursement policy.

SD06.4 Incentive Compatibility in the Delivery of Dialysis Therapy

• Stefanos Zenios; Stanford University, Grad. Sch. of Business, Stanford, CA 94305-5015; stefzen@leland.stanford.edu
• Prashant C. Fuloria; Stanford University, Grad. Sch. of Bus., Stanford, CA 94305; fulhoria_prashant@gsb.stanford.edu

We develop a multi-task principal-agent model that can be used to investigate the impact of different reimbursement policies on the delivery of dialysis therapy. The model assumes that the agent controls the departure rates in a multiclass M/M/infinity queue and the principal reimburses the agent based on the performance of the queue.

Scheduling, Logistics & Facilities Design II

SD07.1 Manufacturing Offshore: What Changes it Brings to your Logistics System

• Russell D. Meller; Auburn University, 308 Dunstan Hall, Dept. of ISE, Auburn, AL 36849-5346; rmeller@eng.auburn.edu
• Pamela H. Vance; Emory University, Goizueta Bus. Sch., 1300 Clifton Rd. NE, Atlanta, GA 30322; pamela_vance@bus.emory.edu

Offshore manufacturing is being utilized or considered by many companies today. A transition to offshore manufacturing has dramatic implications for the logistics system. We discuss a preliminary model for determining which operations should be performed offshore. We also discuss one company's experience with this transition.

SD07.2 Responsibility-Based Design of Networked Production Systems

• Benoit Montreuil; Universite Laval, Operations et Systemes Dec., Pavillon de Seve, Bureau 1152, Sainte-Foy, Quebec, G1K 7P4 , Canada; benoit@osd.ulaval.ca
• Marc Poulin; Universite Laval, NetCentre, Ste-Foy, Quebec, G1K 7P4 , Canada;
• Alain Beaulieu; SET Technologies, Inc., Ste.-Foy, Quebec, G1X 4M8 , Canada;

We present a conceptual framework for developing an integrated design of the operational strategy, the manufacturing organization, the facilities configuration and the manufacturing information architecture of a manufacturing enterprise in terms of a networked production system. We provide an illustrative industrial application.

SD07.3 Integrated Design & Operation of Dual Resource Constrained Flow Shops

• Ronald G. Askin; University of Arizona, Dept. of SIE, PO Box 210020, Tucson, AZ 85721-0020; ron@sie.arizona.edu

Manufacturing cells and flow shops often have more workstations than workers. Cross-trained workers perform multiple operations either with a deterministic orbit for each worker or a more flexible, real-time control strategy. We examine the impact of transfer batch sizes, buffer inventory levels and worker control strategies across a variety of parameterizations of this environment.

SD07.4 Developing Effective Placement Solutions for Indoor Wireless Communications Networks

• Bryan A. Norman; University of Pittsburgh, 1048 Benedum Hall, Dept. of IE, Pittsburgh, PA 15232; banorman@engrng.pitt.edu
• Marty Adickes; University of Pittsburgh, Dept. of IE, Benedum Hall, Pittsburgh, PA 15261;
• Rick Billo; University of Pittsburgh, Dept. of IE, Benedum Hall, Pittsburgh, PA 15261;

Successful installation of indoor wireless communications depends on transceiver placement such that effective radio communication can take place. We present a computerized layout simulation system incorporating Genetic Algorithms as the placement engine for transceivers, coupled with circle covering heuristics to determine initial numbers of transceivers and coverage characteristics.

Game Theoretic Models in OM

SD08.1 A Competitive Modeling of Electricity Pricing

• Chonawee Supatgiat; University of Michigan, Dept. of IOE, 1205 Beal Ave., Ann Arbor, MI 48105-2119; chonawee@umich.edu
• John R. Birge; University of Michigan, Dept. of IOE, 1205 Beal Ave., Ann Arbor, MI 48109-2117; jrbirge@engin.umich.edu
• Rachel Q. Zhang; University of Michigan, 222 IOE Bldg., 1205 Beal Ave., Ann Arbor, MI 48109-2117; rzhang@engin.umich.edu

We consider an open electricity market with uniform pricing rule. The market is modeled as a multi-round auction, which represents a large-scale problem in game theory. The generators face a large decision space and have different generation costs and capacities. We derive their optimal operating policies and the resulting electricity spot prices.

SD08.2 A Stackelberg Game of Pricing

• Hyun-Soo Ahn; University of Michigan, Dept. of IOE, 205 Beal Ave., Ann Arbor, MI 48109-2117; hsahn@umich.edu
• Izak Duenyas; University of Michigan, Dept. of IOE, 1205 Beal Ave., Ann Arbor, MI 48109; duenyas@umich.edu
• Rachel Q. Zhang; University of Michigan, 222 IOE Bldg., 1205 Beal Ave., Ann Arbor, MI 48109-2117; rzhang@engin.umich.edu

We consider a pricing problem faced by a manufacturer who sells his goods at a retail outlet and at its own store. There are 3 pricing decisions to be made: the manufacturer determines the price to sell the product to the retail outlet, the outlet determines its sale price and the manufacturer determines its sale price...

SD08.3 Price Competition under Stochastic Capacity

• Scott Carr; UCLA, Anderson Sch. of Mgmt., Los Angeles, CA;
• Izak Duenyas; University of Michigan, Dept. of IOE, 1205 Beal Ave., Ann Arbor, MI 48109; duenyas@umich.edu
• William S. Lovejoy; University of Michigan, College of Bus., Ann Arbor, MI 48109;

We consider price competition between 2 companies under stochastic capacity. The firms first set their prices which result in orders. They then attempt to produce the products ordered using the (random) capacity they have. There is a penalty for unfilled orders. We model the pricing problem using a game-theoretic model and present structural results.

Group Decision & Negotiation

SD09.1 Electre-GA: A GDSS Based on a Fuzzy Outranking Relation

• Juan C. Leyva-Lopez; Autonomous University of Sinaloa, Facultad de Ingenieria, Ciudad Universitaria, Culiacan, Sinaloa, 80040 , Mexico; jleyva@uas.uasnet.mx

An extension and modification of the ELECTRE III multicriteria outranking methodology is developed to assist a group of decision makers to achieve a consensus on a set of possible alternatives. The procedure is proposed to take place in a specially computer-equipped GDSS room and be conducted by an analyst-facilitator.

SD09.2 An Intelligent Employee Performance Appraisal Writer: An Illustration of Knowledge-Based Group Decision-Making

• Elmore Alexander; Philadelphia College, Sch. House Ln. & Henry Ave., Philadelphia, PA 19144; alexandere@philacol.edu
• Rashmi Malhotra; St. Joseph's University, Dept. of MIS, 5600 City Ave., Philadelphia, PA 19131;
• Davinder K. Malhotra; Philadelphia College, Sch. House Ln. & Henry Ave., Philadelphia, PA 19144; malhotrad@philacol.edu

We illustrate the design and development of a GDSS that uses the capability of knowledge-based expert systems. Although DSS and GDSS are very extensively used by managers, the knowledge management subsystems of current DSSs are not very well developed. To validate and illustrate the usefulness of the MS, we illustrate the development of an intelligent employee appraisal writer...

SD09.3 Joint Gains: Global Negotiation Support in the Internet

• Eero Kettunen; Helsinki University of Technology, Systems Analysis Lab., PO Box 1100, Espoo, HUT, 02015 , Finland; systems.analysis@hut.fi,, http://www.sal.hut.fi
• Raimo P. Hamalainen; Helsinki University of Technology, Systems Analysis Lab., PO Box 1100, Espoo, 02015 , Finland; raimo@hut.fi,, http://www.sal.hut.fi

Joint Gains (www.jointgaines.hut.fi) supports negotiations in the Web using the method of improving directions for multi-party negotiations over multiple continuous issues. All the participants will gain at every iteration. Our other Web-DSS tools and negotiation agents in e-commerce are also discussed.

SD09.4 Division of a Merit Pay Pool by Peers

• Donald N. Stengel; California State University, 5245 North Backer Ave., MS PB07, Fresno, CA 93740-8001; don_stengel@csufresno.edu

A merit pay pool will be divided among members of a department. Departmental peers are the best judges of a colleague's deservedness of merit pay but such an arrangement encourages opportunism by generally underrating others. We propose a mathematically-based procedure for fair division which removes the underrating bias.

Semidefinite Programming & its Applications II

SD10.1 Using SDP for Min-Bisection

• Cris Choi; University of Iowa, Dept. of MS, Tippie Coll. of Business, Iowa City, IA 52242; cchoi@dollar.biz.uiowa.edu
• Yinyu Ye; University of Iowa, Dept. of MS, S384 Pappajohn Bldg., Iowa City, IA 52242-1000; yinyu-ye@uiowa.edu

We present computational results for solving graph min-bisection, i.e., partitioning the nodes of a (nonnegative) weighted graph into two blocks of equal cardinality so as to minimize the weights of crossing edges. In particular, we report computational results on solving a set of circuit partition problems arising in the chip design industry.

SD10.2 A General Framework for Establishing Polynomial Convergence of Long-Step Methods for SDP

• Sam Burer; Georgia Institute of Technology, School of ISyE, Atlanta, GA 30332; burer@math.gatech.edu
• Renato D. C. Monteiro; Georgia Institute of Technology, School of ISyE, Atlanta, GA 30332; monteiro@isye.gatech.edu

We present a general framework for establishing polynomial convergence of long-step methods for SDP. This framework unifies previous polynomial long-step convergence results and is used to establish new polynomial convergence results for long-step methods based on several symmetric central path equations.

SD10.3 An Efficient Algorithm for Solving the Maximum Cut Relaxation SDP Problem

• Renato D. C. Monteiro; Georgia Institute of Technology, School of ISyE, Atlanta, GA 30332; monteiro@isye.gatech.edu
• Sam Burer; Georgia Institute of Technology, School of ISyE, Atlanta, GA 30332; burer@math.gatech.edu

We present an algorithm for solving the positive semidefinite relaxation of the maximum cut problem and compare it to the dual-scaling interior-point algorithm of Benson et al. We present computational experiments to show that our method is quite fast on large scale problems.

Advances in Global Optimization

SD11.1 Global Optimization of Two-Stage Stochastic Mixed Integer Programs

• Shabbir Ahmed; University of Illinois, Dept. of Chemical Eng., 600 South Mathews Ave., Urbana, IL 61801; s-ahmed1@uiuc.edu
• Mohit Tawarmalani; University of Illinois, Dept. of Chemical Eng., 600 South Mathews Ave., Urbana, IL 61801; tawarmal@uiuc.edu
• Nick Sahinidis; University of Illinois, Dept. of Chem. Eng., 600 South Mathews Ave., Urbana, IL 61801; nikos@uiuc.edu

We develop a B&B-based global optimization algorithm for solving 2-stage stochastic MIPs with discretely distributed parameters. We devise specialized lower bounding and branching strategies and discuss finiteness issues. Computational experience with the algorithm is provided.

SD11.2 Portfolio Optimization via Global Nonconvex Quadratic Programming

• J. P. Shectman; University of Illinois, Dept. of Chemical Eng., 600 South Mathews Ave., Urbana, IL 61801; jshectma@uiuc.edu
• Nick Sahinidis; University of Illinois, Dept. of Chem. Eng., 600 South Mathews Ave., Urbana, IL 61801; nikos@uiuc.edu

Markowitz's Nobel-prize-winning oeuvre introduced portfolio optimization with a simple QP model. Empirical evidence mandates a more complicated approach, but computational issues riddle many alternatives. While still quadratic, the proposed method accounts for the qualitative statistical behavior of security returns and investor preferences beyond second order.

SD11.3 Convexification Tools in Integer Programming

• Mohit Tawarmalani; University of Illinois, Dept. of Chemical Eng., 600 South Mathews Ave., Urbana, IL 61801; tawarmal@uiuc.edu
• Nick Sahinidis; University of Illinois, Dept. of Chem. Eng., 600 South Mathews Ave., Urbana, IL 61801; nikos@uiuc.edu

We develop the theory of convex extensions for lower semi-continuous functions and characterize their generating sets through Fenchel transforms. The generating sets are employed to derive tight relaxations for integer programming problems.

SD11.4 An Approach to Parallelize Global Optimization

• Zelda B. Zabinsky; University of Washington, Dept. of IE, Box 352650, Seattle, WA 98195; zelda@u.washington.edu
• Wen Luo; University of Washington, Dept. of IE, Box 352650, Seattle, WA 98195; wendyl@u.washington.edu
• Mark E. Tuttle; University of Washington, Dept. of IE, Seattle, WA 98195; tuttle@u.washington.edu

Advances in parallel computing offer tremendous opportunities in global optimization and raise questions on how to design parallel algorithms in order to make the best use of multiple processors. We will present various implementations of random search methods, with some theoretical bounds and computational experiments on test functions.

Promotion Strategy

SD12.1 Loyalty Rewards Programs: Own or Partnering?

• Atanu R. Sinha; UCLA, The Anderson Sch., 110 Westwood Plaza, Los Angeles, CA 90095-1481; atanu.sinha@anderson.ucla.edu
• Sanjay K. Dhar; University of Chicago, Grad. School of Bus., 1102 East 58th St., Chicago, IL 60637;

Firms in an industry often award rewards on loyalty programs of firms in other industries. Our objective is to examine this phenomenon, both theoretically and empirically, and to determine the conditions under which firms engage in such arrangements and the type of arrangements they use.

SD12.2 Sources & Consequences of Customer Value & Customer Loyalty: Empirical Evidence from the Field

• Indrajit Sinha; Temple University, Fox School of Bus. & Mgmt., Philadelphia, PA; jsinha@sbm.temple.edu
• Wayne S. DeSarbo; Pennsylvania State University;
• Stephanie Young-Helou; ;

While customer value and loyalty programs are becoming increasingly pervasive in many industries and in many countries, empirical evidence linking these constructs in a nomological, behavioral framework to actual consumer purchase behavior is surprisingly scarce. We develop an integrative model of customer value and loyalty involving both their theoretical antecedents and consequences.

SD12.3 Marketing Decisions to Maximize Customer Equity: Application to Promotion Budget Allocation

• Paul D. Berger; Boston University, 595 Commonwealth Ave., Boston, MA 02215; pdberger@bu.edu
• Nada I. Nasr; Boston University, 595 Commonwealth Ave., Boston, MA 02215; nadinasr@bu.edu

We offer a general approach to the promotional budget allocation decision, where the goal is to maximize customer equity. We use decision calculus in which managers' judgments serve as some inputs to formal modeling. A series of applications are offered under different market conditions, including those with synergy, or interaction, between promotional vehicles.

SD12.4 Product Trial & the Disconfirmation Model: A New Perspective

• Gary H. Kritz; Drexel University; gkritz@drexel.edu
• Paul Arsenault; West Chester University, Mktg. Dept., West Chester, PA 19383; parsenault@wcupa.edu

Direct product experience has already been shown to lead to stronger attitude formation. Yet, the marketing literature has very limited research to explain under what conditions product trial works best. We incorporate a well-established model within a product trial framework to begin to help marketers make better promotional dollar allocation decisions.

Marketing Applications of Statistical Models

SD13.1 Crossing the Chasm for High Technology Products: Traditional Diffusion of Innovations or Bi-Logistic Growth

• Lee G. Cooper; UCLA, 330 23rd St., Santa Monica, CA 90402-2514;
• Koen Pauwels; UCLA, Anderson Sch. of Mgmt., 110 Westwood Plaza, Mktg. Dept, Los Angeles, CA 90015-1481;
• Gary J. Russell; University of Iowa, Coll. of Bus. Admin., 108 Pappajohn BAB, Iowa City, IA 52242-1000;
• Wagner A. Kamakura; University of Iowa, Papajohn BAB, Iowa City, IA 52242-1000;

Technology adoption life cycles have traditionally been modeled by the Bass (1969) model of diffusion of innovation. Recently, however, Moore (1994, 1995, and 1998) articulates a different theory of the market adoption of high-technology products. We translate Moore's approach into a model-based framework and estimate it on market data.

SD13.2 A Latent-Class Autologistic Model for Market Based Data

• Anand V. Bodapati; Northwestern University, Kellogg Sch., 2001 Sheridan Rd., Evanston, IL 60208;
• Gary J. Russell; University of Iowa, Coll. of Bus. Admin., 108 Pappajohn BAB, Iowa City, IA 52242-1000;

No abstract supplied.

SD13.3 Predicting & Estimating for the Bayesian NBD Model

• Eric T. Bradlow; University of Pennsylvania, Wharton Sch., Ste. 1400 SH-DH, 3620 Locust Walk, Philadelphia, PA 19104;
• Pete Fader; University of Pennsylvania, Dept. of Mktg., 3620 Locust Walk, Philadelphia, PA 19104-6371;
• Bruce G. S. Hardie; London Business School, Sussex Place, Regent's Park, London, NW1 4SA , England, UK;

No abstract supplied.

Organization Science

Technology & Innovation Management II

SD15.1 How Do Supplier-Buyer Relationships Impact Modularization in NPD?

• Juliana Hsuan Mikkola; Copenhagen Business School, Nansensgade 19, 6, Copenhagen, 1366 , Denmark; jh.ivs@cbs.dk

The degree of modularization, or the opportunity for mixing-and-matching, inherent in a system is dependent upon the number of components and inteface constraints shared among the components, modules and sub-systems. A mathematical modeling approach is used to explain such relationships in addition to how supplier-buyer relationships can impact opportunities for modularization.

SD15.2 Managing the Paradox between Individual Career & Collective Research in Japan: An Organizational Approach

• Jiang Wen; University of Electro-Communications, MS Program, Grad. Sch. of IS, Chofu-ga-oka 1-5-1, Chofu-shi, Tokyo, 182-8585 , Japan; wen@kob.is.uec.ac.jp
• Shinichi Kobayashi; University of Electro-Communications, MS Program, Grad. Sch. of IS, Chofu-ga-oka 1-5-1, Chofu-shi, Tokyo, 182-8585 , Japan;

We provide new observations on managing the paradox between collective research and individual careers in Japan. The evidence, based on university-firm and inter-firm collaborations, shows that such a paradox does not exist in Japan due to the widely existed life-long employment customs and various adopted fluid organizations.

SD15.3 Using Aggregate Process Models to Guide Research & Development Strategy in the Pharmaceutical Industry

• Paul R. Bunch; Eli Lilly & Company, Lilly Corporate Ctr., Indianapolis, IN 46077; bunch@lilly.com
• Aaron L. Schacht; Eli Lilly & Company, Lilly Corporate Ctr., Indianapolis, IN 46077; schacht_aaron_l@lilly.com

We consider the use of aggregate steady-state and dynamic process models to guide R&D strategy in the pharmaceutical industry. The role of models in linking desired R&D output to budgeting, resource allocation and setting performance objectives will be discussed. A case study will be presented that considers resource planning under uncertainty.

SD15.4 Pollution Prevention or Control? A Model of Green Technology Acquisition

• Chialin Chen; University of Illinois, Dept. of Business Admin., 1206 South 6th St., Champaign, IL 61820; cchen3@uiuc.edu,, http://www.students.uiuc.edu/~cchen3
• George E. Monahan; University of Illinois, Dept. of Business Admin., 1206 South 6th St., Champaign, IL 61820; gmonahan@uiuc.edu

A firm's strategic decision concerning green technology acquisition is modeled as a Markov decision process. The focus is on the choice between pollution prevention and control technologies. Key elements of the analysis are the nature of green technology, the decision maker's risk perception and the competitive and regulatory environments.

The Stochastic Programming Approach to Hedge Fund Disaster Protection

SD16.1 The Stochastic Programming Approach to Hedge Fund Disaster Protection

• William T. Ziemba; University of British Columbia, Fac. of Commerce, 2053 Main Mall, Vancouver, BC, V6T 1Z2 , Canada; ziemba@interchange.ubc.ca

Hedge funds are pooled investments for wealthy investors. Macro-funds make directional bets using derivatives. Others take less risk using long and short positions in mis-priced securities. These funds influence I markets with occasionally massive losses and possible bankruptcy. How can such losses be mitigated and prevented using stochastic programming models?

SD16.2 Integrated Financial Risk Management: Capital Allocation Issues

• John M. Mulvey; Princeton University, Program in Financial Eng., Sch. of Eng. & Applied Sci., Princeton, NJ 08544; mulvey@princeton.edu

Optimization models provide the setting for allocating the capital of financial intermediaries. Allocation serves 3 purposes: compares performance across business units, provides a risk-indicator for regulators, develops a basis for major decisions. We show that allocation rules can be retrieved from the solution to a dynamic model of the firm.

SD16.3 Scenario Generation for Multistage Stochastic Programming Portfolio Models in Sequential Sampling Algorithms

• Michael H. Dempster; University of Cambridge, Judge Inst. of Mgmt. Studies, Trumpington St., Cambridge, CB2 1AG , England, UK; mahd2@cam.ac.uk

Sequential sampling procedures for strategic financial portfolio problems relies on the definition of complex scenario generation procedures in which the random behaviour of the considered financial variables is taken into account. We describe a conditional scenario generator, which is interfaced with a recursive model generator and EVPI-importance sampling algorithm.

Cognitive Aspects of Planning & Scheduling Support I

SD17.1 Planning in Dynamic Situations: Some Experience in Process Control & Transportation

• J. M. Hoc; Universite de Valenciennes, Valenciennes, , France; jean-michel.hoc@univ-valenciennes.fr

In cognitive psychology as in AI, studies of planning have been mostly devoted to static situations where the (human) planner fully controls the environment. Consequently, such studies have focused on action plans such as scheduling a series of actions to perform, although integrating hierarchical representations of plans...

SD17.2 Planning in Food Processing Industries

• Wout van Wezel; University of Groningen, Groningen, , The Netherlands; w.m.c.van.wezel@bdk.rug.nl

Market dynamics increasingly pose problems to small- and medium-sized organizations in food processing industries. In particular, customer requirements concerning quality, flexibility and price tend to increase continually. The internal flexibility of food processing organizations must increase in order to respond adequately to market demands. We discuss a planning concept for food processing industries...

SD17.3 Support for Engineering Planning Systems

• Leliane Nunes de Barros; ; leliane@lsi.usp.br

Literature on planning is overwhelming and no clear synthesis of the various planning methods exists that can be used to construct a planner for a particular application. We propose an approach to provide support for engineering planning systems by using a library of problem solving methods. We show how planners can be described in terms of this framework and how they can be conceived as being built from the library conceptual constitutents...

SD17.4 Cognitive Aspect of Plan Execution & Lessons for Overall Planning

• Rene Jorna; University of Groningen;

Supporting the planning task with the help of computers requires extensive knowledge of the execution of this planning task by planners and schedulers. Hayes-Roth & Hayes-Roth (1979) and Garling (1994) studied the cognitive aspects of planning, but only did it with errand-planning tasks. We develop planning and scheduling support systems (Mietus, 1994; van Wezel, Jorna & Mietus, 1996)...

Industrial Applications of Integer Programming

SD18.1 Integer Programming for Scheduling Petroleum Product Distribution

• Steve Welch; Maritrans Management Services Inc., 1818 Market St., Ste. 3540, Philadelphia, PA 19103; swelch@maritrans.com
• Emmanuel D. Chajakis; Princeton University, Princeton, NJ 08544-5263; manos@titan.princeton.edu

We have been using integer programming to schedule petroleum product deliveries across the Gulf of Mexico. As part of a suite of modeling tools, our fleet scheduling MIPs have enabled us to better integrate into our customer's supply chain and optimize the tradeoffs between customer service and fleet utilization.

SD18.2 Complex Industrial (0,1) MIP Problems: Controlling the BB Tree via OSL User Exits - '(Pseudo) Hot Starts'

• Kurt Spielberg; , 38 Knollwood Dr., Cherry Hill, NJ 80002; kurtspie@vnet.ibm.com

We describe a system based on OSL user exits that control the choice of nodes, branch variables and direction. Users can bring their knowledge of the problem, e.g. what variable values are important, to bear directly on the solution process. Applications to '(pseudo) hot starts' will be described.

SD18.3 Location of Emergency Cleanup Equipment for Oil Spill Response Management

• Lefteris Lakovou; University of Miami, Dept. of Indust. Eng., PO Box 248294, Coral Gables, FL 33124; lefteris@eng.miami.edu
• Chi Ip; ;
• Christos Douligeris; University of Miami, Dept. of ECE, PO Box 248294, Coral Gables, FL 33124; christos@ece.miami.edu
• Ashutosh Korde; ;

Myopic planning and tactical decision-making with regard to oil spill events result in severely inefficient cleanup efforts with high fixed, operating and cleanup/damage costs. We propose an integrated framework that addresses some important strategic issues such as determining the optimal location and capacity of cleanup equipment, taking into account their post-event complications...

SD18.4 The Use of Parallel OSL for Refinery Scheduling

• Rob Clark; IBM, 522 South Rd., MSP356, Poughkeepsie, NY 12601; roclark@us.ibm.com
• Gyana Parija; IBM, 522 South Rd., Poughkeepsie, NY 12601;
• Sriram Vasantharajan; Mobil Technology Company, Midway Road, Dallas, TX 75244; sriram_vasantharajan@email.mobil.com
• Ramesh Raman; Mitsubishi Chemical, 99 Tasman Dr., San Jose, CA 95134;

We will discuss our experiences with solving a difficult problem of planning the development and operation of oil fields over a specified time horizon. Several MIP strategies were developed in the GAMS/OSL integrated solver and applied to the GAMS model. Results from the parallel GAMS/OSL solver will be given.

Telecommunications

SD19.1 Planning Cordless Business Communication Systems

• Pascal Brisset; University for Civil Aviation, 7 ave. Edouard Belin, Toulouse Cedex, 31055 , France; brisset@recherche.enac.fr
• Thom Fruehwirth; Ludwig-Maximilians University, Oettingenstr. 67, Munich, D-80538 , Germany; fruehwir@informatik.uni-muenchen.de

We describe a new tool to optimize the placement and minimize the number of base stations for local wireless communications at company sites. It uses high-level constraint-based technology on top of a logic programming language. This example shows that efficiency is not incompatible with concise and elegant programming.

SD19.2 Radio Link Frequency Assignment Benchmark

• Thomas Schiex; INRA, Biometry & AI Dept., Chemin de Borde Rouge BP27, Castanet-Tolosan, 31326 , France; tschiex@toulouse.inra.fr
• Bertrand Cabon; ONERA-CERT, Comp. Science Dept., 2 ave. Edouard Belin, Toulouse Cedex, 31025 , France; bertrand.cabon@tls.mms.fr
• S. de Givry; ONERA-CERT, Comp. Science Dept., 2 ave. Edouard Belin, Toulouse Cedex, 31025 , France; degivry@thomson-lcr.fr
• L. Lobjois; ONERA-CERT, Comp. Science Dept., 2 ave. Edouard Belin, Toulouse Cedex, 31025 , France; lobjois@cert.fr
• Joost P. Warners; Delft University of Technology, Dept. of Tech. Math & Info., Zuidplantsoen 4, Delft, 2628 BZ , The Netherlands; joost.warners@cwi.nl

The problem of radio frequency assignment is to provide communication channels from limited spectral resources while keeping to a minimum the interference suffered by those wishing to communicate in a given radio communication network. We will present a set of RLFAP benchmark with some results, including how constraint satisfaction techniques can cope with these problems.

SD19.3 Analyzing & Planning Traffic Flows for IP Service

• Robert Rodosek; Imperial College, IC-PARC, William Penney Lab., London, SW7 2AZ , England, UK; r.rodosek@icparc.ic.ac.uk

The task of efficiently managing spare bandwidth capacity in the network is imperative for ISPs. We present a constraint-based approach to analyze and plan traffic flows in case of changes, e.g., peering. The existing flows are recognized up to a required precision by reasoning on minimum network data.

SD19.4 An Optimal Strategy for the Constrained Cycle Cover Problem

• Patrick Soriano; University of Montreal, HEC, CP 6128, succ. Centre-ville, Montreal, Quebec, H3C 3J7 , Canada; partick@crt.umontreal.ca
• Gilles Pesant; University of Montreal, CRT, CP 6128, Succ. Centre-ville, Montreal, Quebec, H3C 3J7 , Canada; pesant@crt.umontreal.ca

We propose a constraint programming algorithm for the decomposition of a telecommunications network into a set of rings in order to protect it against failures. Optimal solutions are obtained efficiently on networks of realistic size.

Traveler Behavior Models

SD20.1 Temporal-Spatial Micro-Assignment of Travel Demand with Activity Chains

• Ahmed F. Abdelghany; University of Texas, Dept. of Civ. Eng., ECJ 6.2, Austin, TX 78712;
• Hani S. Mahmassani; University of Texas, Dept. of Civil Eng., ECJ 6.2, Austin, TX 78712-1076;

The assignment problem of travel demand with activity chains is presented. A stochastic dynamic user equilibrium solution algorithm is developed and implemented incorporating a traffic simulation model within an iterative search framework. Experiments conducted on a test network provide insight on system performance under travel demand with activity chains.

SD20.2 A Utility-Theory-Consistent System-of-Demand Equations Approach to Household Travel Choice

• Kara Kockelman; University of Texas, Dept. of Civ. Eng., ECJ 6.9, Austin, TX 78712; kkockelm@mail.utexas.edu

A utility-theory-consistent system of demands for activity participation in a time-price setting is derived. A time budget, not just a financial one, and the discrete nature of many travel-related demands are incorporated explicitly. A multivariate negative binomial stochastic specification produces closed-form observation likelihoods and permits unobserved heterogeneity across households, along with correlation within a single household's set of demands.

SD20.3 Investigation of Processes Underlying Commuters Dynamic Route & Departure Time Choices under Real-Time Information

• Karthik Srinivasan; University of Texas, Civil Eng., ECJ 6.2, Austin, TX 78712;
• Hani S. Mahmassani; University of Texas, Dept. of Civil Eng., ECJ 6.2, Austin, TX 78712-1076;

Mechanisms underlying commuter behavior dynamics including learning and adjustment processes are investigated. This research has important implications for accurate traffic modeling and network state prediction and design of effective control strategies.

SD20.4 User Behavior under Variable Message Signs

• Srinivas Peeta; Purdue University, Sch. of Civil Eng., West Lafayette, IN 47907;
• Jorge Ramos; Purdue University, Sch. of Civ. Eng., West Lafayette, IN 47907-1284;

Models are developed to predict diversion rates on highways based on the content of messages displayed. The Borman Expressway in Northern Indiana is used as a case study.