Activity Number:
|
364
|
Type:
|
Contributed
|
Date/Time:
|
Wednesday, August 14, 2002 : 2:00 PM to 3:50 PM
|
Sponsor:
|
Biopharmaceutical Section*
|
Abstract - #300914 |
Title:
|
A Goodness-of-fit Test for Overdispersed Logistic Regression Model
|
Author(s):
|
Santosh Sutradhar*+ and Nagaraj Neerchal and Jorge Morel
|
Affiliation(s):
|
Merck & Company, Inc. and University of Maryland, Baltimore County and Procter & Gamble Company
|
Address:
|
BLX-27, 518 Township Line Road, Blue Bell, Pennsylvania, 19422, USA
|
Keywords:
|
Cluster count data ; Goodness-of-fit test ; Logistic regression ; Maximum likelihood estimation ; Overdispersion models ; Parametric bootstrapping
|
Abstract:
|
In many practical situations, we observe binomial count data coming from clusters along with a set of measurements on covariates associated with each cluster. Due to clustering, these binomial counts exhibit larger variances than that permitted by the binomial model. This phenomenon is know as overdispersion or extra-variation. Sutradhar (2001), in his PhD dissertation, discussed a goodness-of-fit (GOF) test for testing for the model adequacy of overdispersed binomial distribution. This statistic is a direct analogue of the usual Pearson Chi-square statistic. It can be used for choosing an appropriate model for a given data set from several competing overdispersion models. The test is also applicable to the situation where the cluster sizes are not necessarily equal. In this presentation, we illustrate the use of this GOF-test in a general linear model with logistic link, using a data set from a study of a teratology experiment designed to investigate the synergistic effect of the anticonvulsant phenytoin and trichloropropene oxide on the parental development of inbred mice.
|
- The address information is for the authors that have a + after their name.
- Authors who are presenting talks have a * after their name.
Back to the full JSM 2002 program |