|
Activity Number:
|
26
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Sunday, July 29, 2007 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
General Methodology
|
| Abstract - #309747 |
|
Title:
|
Iterative Modified Likelihood Ratio Test for Homogeneity
|
|
Author(s):
|
Pengfei Li*+ and Jiahua Chen
|
|
Companies:
|
University of Waterloo and University of British Columbia
|
|
Address:
|
Unit 305, Waterloo, ON, N2L6P4, Canada
|
|
Keywords:
|
Asymptotic distribution ; Exponential mixture ; Finite mixture model ; Likelihood ratio test ; Score test
|
|
Abstract:
|
Testing for homogeneity in finite mixture models has attracted substantial research recently. Modified likelihood ratio test (MLRT) is a nice method because it has an asymptotically distribution-free test statistic and is locally most powerful. Interestingly, the mixture of exponential distributions or mixture models in scale distribution families do not satisfy the regularity conditions prescribed by many methods including the MLRT. To overcome this difficulty, we propose an iterative modified likelihood ratio test (IMLRT) in this paper. The IMLRT statistic has the same simple limiting distribution as MLRT statistic. The result is applicable to much more general mixture models and it does not require the parameter space to be bounded. Simulations show that the IMLRT has more accurate type I errors and higher powers under various models compared to existing methods.
|
