AN INTRODUCTION TO BAYESIAN MODEL SELECTION

Many data analysis problems in science reduce to one of model selection. For example, a materials scientist investigating the structure and properties of a new compound may wish to know: are there two peaks under these diffraction data or three? For how many layers of materials is there most evidence in these reflectivity measurements? Do these data indicate a phase transition? In this introductory presentation, we show how probability theory can be used to address such questions in a simple and straight forward manner.