Sankhya: The Indian Journal of Statistics

1998, Volume 60, Series B, Pt. 1, 145--160

BAYESIAN MODELLING OF A PHARMACEUTICAL EXPERIMENT WITH HETEROGENEOUS RESPONSES

By

CHRISTIAN P. ROBERT, Laboratoire de Statistique, CREST, Paris
Université de Rouen, France
and
ANNIE REBER, Université de Rouen, France

SUMMARY. We show in this paper how a hierarchical Bayes modelling can lead to better insights in a pharmaceutical experiment assessing the performances of a drug against vertigo, by providing direct access to different extensions of the model. The abailability of such extensions relies on the use of Gibbs sampling techniques which allow for sarbitrary distributions. The Bayesian processing of the experiment is noninformative, which allows for an objective perspective, and we consider in detail the extension when the population is heterogeneous at some stage of the experiment, and the corresponding modelling by a normal mixture distribution.

AMS (1991) subject classification. 62F15, 62J10,62M99, 62P10.

Key words and phrases. Analysis of variance; Gibbs sampling; missing data; mixtures; normal linear model; placebo effect; probit model.

Full paper (PDF)

This article in Mathematical Reviews