PRAKASH LAUD
Division of Biostatistics
Medical College of Wisconsin

ABSTRACT

Over the past few decades, many authors have proposed predictive methods for selecting a model from a pool of models. In this talk we first briefly review these and then focus on the developments of the recent few years. In particular, we consider methods based on the predictive density of a replicate experiment that provide several criteria for model selection. A salient feature of these methods is the provision of a measure of uncertainty in model choice. Application of these methods to linear, generalized linear and survival models is discussed.