|
Activity Number:
|
336
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, July 31, 2007 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
IMS
|
| Abstract - #308490 |
|
Title:
|
Testing Homogeneity Variances for Non-Normal Models Using Entropy
|
|
Author(s):
|
Solomon W. Harrar*+ and Arjun K. Gupta and Leandro Pardo
|
|
Companies:
|
University of Montana and Bowling Green State University and Complutense University of Madrid
|
|
Address:
|
32 Campus Drive, Missoula, MT, 59812,
|
|
Keywords:
|
Sample Variance ; Cumulants ; Edgeworth Expansion ; Characteristic Function ; Robustness ; Entropy
|
|
Abstract:
|
This talk is concerned with comparison of variability among populations which are not necessarily normally distributed. Our measure of variability is based on an information theoretic criteria known as Shannon's Entropy. The main problem with the standard methods is that they are sensitive to the distributional assumptions. Based on Shannon's Entropy, we derive a test for equality of variability assuming a general location scale family for the populations. Asymptotic expansion is obtained for the null distribution of the test statistic. The technique develops a multivariate Edgeworth expansion of the characteristic function and a formal inversion of the expanded characteristic function. Our test is shown to perform well compared to the likelihood ratio criterion when the populations have normal as well as non-normal distributions.
|
