Academic Press Library in Signal Processing

Chapter 25

A Tutorial on Model Selection

Enes Makalic*, Daniel Francis Schmidt* and Abd-Krim Seghouane^†,    ^*Centre for M.E.G.A. Epidemiology, The University of Melbourne, Australia,    ^†Department of Electrical and Electronic Engineering, The University of Melbourne, Australia

Abstract

Model selection is a key problem in many areas of science and engineering. Given a data set, model selection involves finding the statistical model that best captures the properties and regularities present in the data. This chapter examines three broad approaches to model selection: (i) minimum distance estimation criteria, (ii) Bayesian statistics, and (iii) model selection by data compression. Model selection in the context of linear regression is used as a ...