|
Activity Number:
|
475
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 9, 2006 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Statistical Computing
|
| Abstract - #306114 |
|
Title:
|
Factor Analysis for Multiattribute Ranked Data
|
|
Author(s):
|
Philip L. H. Yu*+ and Wai Ming Wan
|
|
Companies:
|
The University of Hong Kong and The University of Hong Kong
|
|
Address:
|
Pokfulam Road, Hong Kong, 00852, China
|
|
Keywords:
|
ranked data ; factor analysis ; MCEM
|
|
Abstract:
|
This paper develops factor models for ranked data in which items are ranked based on several attributes or criteria. In modeling multi-attribute ranking data, two sources of item-response dependencies have to be distinguished. Within-attribute dependence arises when items are evaluated on the same attribute and between-attribute dependence emerges when items are compared with respect to different attributes. We extend the factor model proposed by Yu, Lam and Lo (2005) for single-attribute ranked data so that it takes into account the between-attribute dependence as well. The Monte Carlo Expectation Maximization (MCEM) algorithm is used for parameter estimation. A bootstrap method is proposed for assessing the fitness of a model. Simulation studies are carried out to demonstrate the proposed estimation and goodness-of-fit methods.
|
