|
Activity Number:
|
327
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Tuesday, July 31, 2007 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Nonparametric Statistics
|
| Abstract - #309122 |
|
Title:
|
A New Hypothesis-Testing Approach When the Nuisance Parameter Is Only Present Under the Alternative Hypothesis
|
|
Author(s):
|
Olivier Thas*+ and Gang Zheng
|
|
Companies:
|
Ghent University and National Heart, Lung, and Blood Institute
|
|
Address:
|
Coupure Links 653, Gent, 9000, Belgium
|
|
Keywords:
|
hypothesis testing ; nuisance parameters ; nonparametric test ; goodness-of-fit ; empirical process ; genetics
|
|
Abstract:
|
We present a new class of statistical tests. The statistic is defined as the length of an excursion set (LESS) of an empirical process. Examples of testing situations where the test may be useful include goodness-of-fit, linkage analysis, and testing a hypotheses when a nuisance parameter is present only under the alternative. We give the asymptotic distributions under both the null hypothesis and under contiguous alternatives. By defining an excursion as an exceedance of an upper or lower boundary function, these functions may be chosen to get good power. In a simulation study, we empirically investigate the behavior of the tests, and we illustrate its application to the examples mentioned earlier.
|
