An improved method for model selection based on Information Criteria

Abstract

Information criteria are an appropriate and widely used tool for solving model selection problems. However, different ways to use them exist, each leading to a more or less precise approximation of the sought model. In this paper, we mainly present two methods of utilisation of information criteria : the classical one which is generally used and an alternative one, more precise but requiring a little more calculations. Those methods are compared on 1-D and 2-D autoregressive models ; we use a synthetized process for the 1-D case and texture images for the 2-D case. We also work with the original phi_beta criterion which includes all others usual criteria such as AIC, BIC, and phi.