|
Activity Number:
|
314
|
|
Type:
|
Invited
|
|
Date/Time:
|
Tuesday, July 31, 2007 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Statistics and Marketing
|
| Abstract - #308047 |
|
Title:
|
Unobserved Change in Conjoint Analysis
|
|
Author(s):
|
Thomas Otter*+ and Sylvia Fruehwirth-Schnatter and Regina Tuechler
|
|
Companies:
|
The Ohio State University and Johannes Kepler Universitaet Linz and Vienna University of Economics and Business Administration
|
|
Address:
|
Department of Marketing & Logistics, Columbus, OH, ,
|
|
Keywords:
|
conjoint analysis ; change point ; heterogeneity ; Bayesian Analysis
|
|
Abstract:
|
We develop models of change that are parsimonious in the sense that each respondent is characterized by one vector of regression weights, like in the standard mixed effects model. In the first model every respondent's observation error variance is allowed to change according to a change point process. The change points are modeled as random effects. In the second model change is viewed to affect which attributes are being used by a respondent in forming product evaluations. This may be expressed as a dynamic variable selection problem where indicators pointing to the active variables evolve over the course of the experiment. We model these indicators as random effects that evolve according to a change point process. We conclude with a comparison to an approach where observations separated by change points are independent (conditional on hyper-parameters).
|
