Analysis of clustered ordinal data with subclusters via a bayesian hierarchical model

Abstract

We propose a Bayesian hierarchical model to analyze clustered ordinal data. We describe the method in the context of a clinical trial for examining the efficacy of tretinoin emollient cream in the treatment of photo-aged skin. For each patient there were five ordinal responses (each with 4 ordinal categories) from two locations, four from arms and one from face. Thus the data were nested with two levels: patient level and location level. The Bayesian hierarchical model gives estimates for location-specific treatment effect and credible intervals.