From shx@psu.edu Mon Sep 20 11:44:15 2004 Received: from f04n07.cac.psu.edu (f04s07.cac.psu.edu [128.118.141.35]) by math.psu.edu (8.12.10/8.12.10) with ESMTP id i8KFiEkC008624 for ; Mon, 20 Sep 2004 11:44:15 -0400 (EDT) Received: from sxu-laptop.psu.edu (net06pc71.smeal.psu.edu [146.186.139.71]) by f04n07.cac.psu.edu (8.12.11/8.12.11) with ESMTP id i8KFhhc8190016; Mon, 20 Sep 2004 11:43:43 -0400 Message-Id: <6.1.0.6.2.20040920114430.033da728@mail.psu.edu> X-Sender: shx@mail.psu.edu X-Mailer: QUALCOMM Windows Eudora Version 6.1.0.6 Date: Mon, 20 Sep 2004 11:45:44 -0400 To: "Rich Andrew" , cea@psu.edu, sfa@psu.edu, "Russell Barton" , byrne@games.arl.psu.edu, Enrique del Castillo , tmc7@psu.edu, mjc3@psu.edu, k1c@psu.edu, cti@psu.edu, Sunmee Choi , fyj@psu.edu, "Lily Elefteriadou", eee1@psu.edu, eur@psu.edu, fa2@psu.edu, i2v@psu.edu, mbf10@psu.edu, nxg11@psu.edu, rlg3@psu.edu, mch@psu.edu, c1h@psu.edu, tharrison@psu.edu, jch@psu.edu, hzh@psu.edu, kesidis@engr.psu.edu, j9l@psu.edu, hsl2@psu.edu, g5l@psu.edu, costas@psu.edu, mpt@psu.edu, j93@psu.edu, DPassmore@psu.edu, rvr@psu.edu, axr32@psu.edu, "Doug Thomas" , "William Rothwell", mcs5@psu.edu, jss15@psu.edu, "Timothy Simpson", ttc@psu.edu, jvt@psu.edu, "Leonid Vaserstein", jav1@psu.edu, pnw@psu.edu, avw@psu.edu, shx@psu.edu, edm3@psu.edu, bxy121@psu.edu, mlb336@psu.edu, sib105@psu.edu, jdc220@psu.edu, clc16@psu.edu, cxc505@psu.edu, sac164@psu.edu, vkd106@psu.edu, amd287@psu.edu, sud111@psu.edu, jif1@psu.edu, jrf163@psu.edu, lug113@psu.edu, gmm17@psu.edu, auj106@psu.edu, vmk111@psu.edu, kwanisai@stat.psu.edu, zilong@psu.edu, ananth@psu.edu, sym110@psu.edu, hgm106@psu.edu, mxm74@psu.edu, mxo203@psu.edu, aum144@psu.edu, sem203@psu.edu, nhprasad@psu.edu, mxm74@psu.edu, sxo151@psu.edu, asp118@psu.edu, jkr143@psu.edu, agr119@psu.edu, rvr4@psu.edu, bqr2@psu.edu, nazrul@psu.edu, npt103@psu.edu, kst126@psu.edu, kxt178@psu.edu, sft108@psu.edu, axt206@psu.edu, oav100@psu.edu, nuv104@psu.edu, ejw169@psu.edu, lxw167@psu.edu, yxw133@psu.edu, mcw150@stat.psu.edu, yxw146@psu.edu, sha1@psu.edu, yus100@psu.edu, mbz113@psu.edu, yxz11@psu.edu From: Susan Xu Subject: OR Colloquium, Sept 21. Speaker: Professor Duncan Fong Mime-Version: 1.0 X-Virus-Scanned: by amavisd-new X-Virus-Status: No X-Spam-Checker-Version: SpamAssassin 2.60 (1.212-2003-09-23-exp) on leibniz.math.psu.edu X-Spam-Level: X-Spam-Status: No, hits=-0.0 required=5.0 tests=BAYES_44 autolearn=ham version=2.60 X-Lines: 65 Status: RO Correction: The presentation date of Dr. Fong's seminar should be Sept 21, 2004. Thanks. Susan OR 590 COLLOQUIUM 4:15-5:30 PM, September 21, 2004 109 Walker Building Dr. Duncan K. H. Fong Professor of Marketing and Professor of Statistics Penn State University Bayesian Methodology for Detecting and Estimating Regime Change Points and Variable Selection in Multiple Regression Models for Marketing Research Abstract. Multiple regression is one of the most frequently utilized multivariate statistical procedures applied in the Marketing Research arena for examining the impact of one or more independent variables upon a designated dependent variable. The estimated regression coefficients denote the contribution of changes in the levels of the independent variables upon the dependent variable. With time series or longitudinal data, these contributions are typically assumed constant over the entire data collected for the particular study at hand. However, as is known in many such Marketing applications, the effects of such independent variables may change over the course of the time period investigated. Also, in many surveys involving repeated measures, fatigue, order effects, learning, patterned responses, etc. may also lead to alterations in the contributions of these independent variables over the sequence of administration. We present a Bayesian change point multiple regression methodology for such studies which, for a fixed number of change points, simultaneously estimates the location of change points in time/sequence, the corresponding relevant subset of independent variables per regime, and the associated regression parameters of these selected independent variables per regime. Furthermore, a probability based model selection heuristic, the Bayes factor, is derived in this particular modeling context to determine the appropriate number of change points in the data. An application involving actual time series data involving prescription sales for an ethical drug is utilized to demonstrate the methodology. Finally, we discuss directions for future research. Biography: Duncan K. H. Fong is Professor of Marketing and Professor of Statistics at The Pennsylvania State University where he has been on the faculty since 1987. He holds a Ph.D. in statistics from Purdue University. His research interests include marketing research, Bayesian statistics, experimental economics, forecasting and supply chain management. He is currently working on several projects, which include the development of new methodologies to perform conjoint analysis, the investigation of dynamic evolution of consumer preferences, the study of regime changes with an application to multiple regression analysis, among others. He has been active in research and is the author or co-author of more than thirty articles published in various academic journals. Professor Fong is an Associate Editor of ISBA Bulletin, the official publication of the International Society for Bayesian Analysis, from March 2001 to June 2004. He is also a Founding member of the International Society for Bayesian Analysis and served on the Board from 1992 to 1994. He has supervised four doctoral dissertations, one master thesis and one undergraduate honors thesis. Furthermore, he has served on seventeen Ph.D. and twelve MS students' committees. He teaches business statistics and marketing research courses to undergraduate, MBA, M.S. and Ph.D. students.