A Bayesian Analysis of Heterogeneity in Ordinal Regression
by
Samuel O.M. Manda
Green Lane Hospital, Auckland Healthare Ltd, Auckland, New Zealand

A Bayesian methodology is developed to measure the degree of heteogeneity in ordinal regression. The methodology assumes a logit link function for the cumulative probabilty of an ordinal response level. The methodology can be applied to treatment effects in a multi-center clinical trial where the response is ordinal. If there is homegeneity of treatment effects, then the conclusions concerning those effects apply to wider population of patiens. Otherwise, heterogeneity and its possible effects on the trial need to be explained. The methodology is tested on a previously analyzed (using the EM) social image data from Taiwan. The main example data come from a multi-center clinical trial on therapies for acute myocardial infarction.

Date received: May 1, 2000

Copyright © 2000 by the author(s). The author(s) of this document and the organizers of the conference have granted their consent to include this abstract in Atlas Conferences Inc. Document # cadt-04.