Fall of 2006
Welcome to the Management Science and Operations Management (MSOM) Seminar at the School of Business Administration at the University of Miami.
Here you will find details about the talks given during the Fall of 2006.
Past Talks
- December 11th
Speaker: Ling Wang, Department of Operations Management, University of Miami
Title: Spline-Backfitted Kernel Smoothing of Additive Models in Time Series
Abstract: Application of non- and semi parametric regression techniques to high dimensional time series data have been hampered due to the lack of effective tools to address the ''curse of dimensionality''. Under rather weak conditions, we propose a spline-backfitted kernel estimator of the component functions for the nonlinear additive time series data that is both computationally expedient so it is usable for analyzing very high dimensional time series, and theoretically reliable so inference can be made on the component functions with confidence. Simulation experiments have provided strong evidence that corroborates with the asymptotic theory. Finally, the estimation procedure has been illustrated by a gas furnace example.
- December 8th
Speaker: Huiliang Xie, University of Iowa
Title: Extending SCAD-Penalized Regression to Accelerated Failure Time Models
Abstract: In linear regression, variable selection becomes important when a large number of predictors are present and it is believed that the majority of these predictors are not important. The high-dimensional problem is further complicated when the response variables may be right-censored, as is a common case in survival analysis. We investigate regression with the smoothly clipped absolute deviation (SCAD) penalty in the accelerated failure time (AFT) model and proposed the SCAD-penalized AFT regression. We showed that the method defined can automatically perform consistent variable selection and at the same time estimate the coefficients associated with the important predictors as efficiently as if the trivial predictors had been ruled out from the very beginning. In our study, we allow the number of predictors pn to go to infinity as the number of observations n goes to infinity. This study is the first rigorous study of penalized approaches assuming the AFT model. The majorize-minimize algorithm is adapted to compute the SCAD-penalized Kaplan-Meier weighted least squares in this set-up. Finite sample behavior of this estimator is studied via simulation and demonstrated by two data examples.
- December 5th
Speaker: Xuemei Shan, Northwestern University
Title: Blind Discovery of Manufacturing Variation Sources
Abstract: In modern manufacturing processes, large quantities of multivariate data are routinely available through automated measurement and sensing. The data generally contain a great deal of buried diagnostic information that can aid in identifying and eliminating major root causes of manufacturing variation. Traditional statistical process control tools, which were designed for situations in which data are much less abundant, do not take full advantage of in-process measurement. In this talk, I will discuss approaches that are much better suited for this scenario, which involve blindly identifying the number of sources that are present and the spatial and temporal pattern of each source. Here, "blind" refers to the fact that no prior knowledge of the variation sources or their patterns is assumed. Examples will be used to illustrate the proposed methods.
- November 1st
Speaker: Ted Klastorin, Department of Information Systems and Operations Management, University of Washington Business School
Title: On the Introduction of New Products in a Monopoly IT Market
Abstract: In this paper, we investigate the existence of an inter-temporal price discriminating equilibrium for a rapidly innovating good, such as software, produced by a monopolist. We assume that the monopolist is planning to introduce a new IT product that may have multiple upgrades or versions during the finite life span of the product. Given the nature of the product (software), there is no variable production costs although there are development costs that increase directly as a function of the design/quality of the product, where the design/quality of the product can be measured by a single scalar. Following previous research, we assume that consumers enter the market over time according to a modified Bass-type diffusion process. After entering the market, consumers purchase a product if their utility surplus is positive. Consumers who purchase the first version are offered an upgraded version at a discounted price. The monopolist must set the level of design/quality and price for each version as well as the timing of each new upgrade introduction. In this paper, we develop a model to analyze this product introduction problem when only two versions are considered (i.e., one upgrade). Our model suggests a number of important insights, including the finding that the price offered to previous purchasers to upgrade should (almost always) be less than the price for new purchases of the upgraded version. In addition, we prove an important relationship between the optimal revenue and development cost for monopoly markets, when development costs are defined as a quadratic function of product quality. We illustrate and discuss a number of other significant managerial implications based on extensive numerical analyses.
- October 13th
Speaker: Paulo Gonçalves, Department of Management Science, University of Miami
Title: Evaluating Overreaction to Backlog as a Behavioral Cause of the Bullwhip Effect
Abstract: We evaluate, in an experiment with the Beer Distribution Game, a complementary behavioral source of the bullwhip effect that has been previously ignored in the literature: overreaction to backlogs. By separating the estimation of the response to inventory and backlog, we find that players treat backlog differently than inventory. Contrary to our expectations, players do not over-order when in backlog; instead, they have a measured response, saturating order adjustment and limiting the amount of amplification they introduce in the order stream. This result is robust across echelons, model specifications, and data sets. By structuring data from 25 games as a panel, we aggregate data across individuals and echelons, and improve the representativeness of the estimated decision rules. Whereas previous research suggests that the supply line is under-accounted for, we find that most players ignore it and that only players facing short and uncapacitated supply lines are capable of taking it into consideration. We also find that inventory adjustment is more aggressive in upstream echelons of the supply chain, where players face higher order variance. Still, players across the supply chain maintain similar levels of desired inventory. Since under-reacting to shortages is stabilizing and not accounting for the supply line is destabilizing, performance of the estimated policy depends on the relative contribution of individual components. Given the information cues available, policy performance suggests that players show bounded rationality and develop a "satisficing" replenishment decision rule that minimizes local cost at the expense of higher upstream cost. We explore the implications of these findings for the design of information and incentive systems for supply chain management.
- September 20th
Speaker: Paulo Gonçalves, Department of Management Science, University of Miami
Title: Behavioral Causes of Product Returns in the Seed Supply Chain
Abstract: The hybrid seed industry experiences excessive and costly rates of seed returns from distributors. Dealers must order in advance of grower demand realization, and may return unsold seed to manufacturers at the end of the season. Returns impose substantial costs on the manufacturer, including transportation, conditioning, discards and excess production capacity. Here we investigate the processes that lead to high return rates through a model-based field study. We develop a formal dynamic model of sales resource allocation as it interacts with customer and producer behavior in the seed supply chain. While sales representatives know they should carefully position seed types and quantities to match heterogeneous demand by learning about dealer and grower needs, we show, however, that sales agents abandon time-consuming positioning to meet quotas late in the sales cycle. Such sales "push" leads to excessive returns in the next period, increasing the total sales agents must attain to reach their quota, leading them to push still more seed. Model analysis shows how this positive feedback can tip the system into a self-reinforcing high-return equilibrium. We discuss policies to reduce returns and implementation issues that arise. The implications are of general interest because the seed industry is similar to other high-technology, high-velocity industries characterized by short and unpredictable product lifecycles, rapid turnover of SKUs in the catalog, long product development and production delays, and volatile and unpredictable customer demand.