|
Activity Number:
|
320
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 5, 2008 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Nonparametric Statistics
|
| Abstract - #300523 |
|
Title:
|
Tests of Fit for the Beta Binomial Distribution
|
|
Author(s):
|
John Best*+ and John Rayner and Olivier Thas
|
|
Companies:
|
University of Newcastle, Australia and University of Newcastle, Australia and Ghent University
|
|
Address:
|
12 Tobruk Ave, Carlingford, 2118, Australia
|
|
Keywords:
|
Goodness of fit ; Moment tests ; Empirical distribution function ; Discrete distributions ; Power studies
|
|
Abstract:
|
The beta-binomial has recently been suggested as an alternative model for analysis of simple market research preference tests (Bi and Ennis, 2001). Before accepting its use for this application and others it is sensible to test the fit of the model. Traditionally this can be done with a Pearson chi-squared test. Simonoff (2003) suggested a dispersion test while Lockhart et al (2007) discuss general tests of fit for discrete distributions based on the empirical distribution function. Skewness and kurtosis tests of fit will also be examined here. Preliminary results show the dispersion test has poor power for some alternatives, the Pearson test has some power for most alternatives, either the skewness or kurtosis test often has good power while the Anderson-Darling test is generally more powerful than the other tests for the alternatives considered.
|
