|
Activity Number:
|
425
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 1, 2007 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Statistics and Marketing
|
| Abstract - #309208 |
|
Title:
|
A Bayesian Approach for the Multidimensional Scaling of Ordered Preference Data
|
|
Author(s):
|
Duncan K.H. Fong*+
|
|
Companies:
|
The Pennsylvania State University
|
|
Address:
|
456 Business Building, University Park, PA, 16802,
|
|
Keywords:
|
Multidimensional Scaling ; Bayesian Analysis ; Preference Analysis
|
|
Abstract:
|
We develop a Bayesian MDS vector model to analyze ordered preference data. Unlike classical MDS procedures, the Bayesian method includes a probability based criterion to determine the number of dimensions of the derived joint space map. Also, our procedure models the raw data which ameliorates the need of any data preprocessing as required for some widely used classical MDS procedures. A unique feature of the proposed Bayesian procedure is that it allows external attribute information to be directly incorporated into the spatial representation of the preference data, and that the derived prior eliminates the need of identification constraints on the bilinear structure of the model to obtain a proper posterior distribution. We solve the computational problem associated with the use of an intractable posterior distribution from one data set as a prior for the analysis of another dataset.
|
