|
Activity Number:
|
475
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 9, 2006 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Statistical Computing
|
| Abstract - #306914 |
|
Title:
|
Assessment of Influential Observations Using Alpha Factor Analysis
|
|
Author(s):
|
Zenaida F. Mateo*+ and Yutaka Tanaka
|
|
Companies:
|
University of Manitoba and Nanzan University
|
|
Address:
|
Department of Statistics, Winnipeg, MB, R3T 2N2, Canada
|
|
Keywords:
|
alpha factor analysis ; influence functions ; Euclidean norms ; common factors ; unique variance ; common factor decomposition
|
|
Abstract:
|
Alpha Factor Analysis (AFA) is one of the popular methods of factor analysis which was proposed based on the psychometric concept of generalizability. It's basic idea is to determine the common factor (fj) in such a way that they have maximum correlation with the corresponding universe common factors. The main objective in this study is to evaluate the influence of a small change of data on the values of the unique variance matrix and the common variance matrix. This was analyzed by deriving some theoretical influence functions of the common variance matrix and the unique variance matrix respectively in the common factor decomposition using perturbation theory. To assess the influential observation, some influential measures like the Euclidean norms were utilized using the proposed procedure. Some examples and comparisons will be presented to illustrate the present procedure.
|
