Contrasting Bayesian analysis of survey data and clinical trials

Although both surveys and clinical trials are amenable to Bayesian hierarchical modelling, the general aims, constraints and actual analysis of each can often vary considerably. First, examples are presented showing how Bayesian hierarchical modelling can be used to produce estimates for small areas from survey data and, also, how it can be used to combine data from clinical trials. Then, it will be shown how surveys and clinical trials may differ with respect to the presence of design effects/selection biases and with the ability to validate models. The impact of the design on modelling will be highlighted and a class of sample selection models will be shown to help alleviate the design's influence. Although surveys generally have enough data to validate many features of a model, clinical trials may not, leaving sensitivity analysis as a means to prior acceptance. Some design issues, contrasting Bayesian with frequentist methods, will also be discussed. Published in 2001 by John Wiley & Sons, Ltd.