BS2 Further Statistical Inference HT08 - Lectures 9 and 10
Model comparison and determination
This lecture discusses issues and criteria for model comparison and determination, in particular two Akaike's Information Criterion (AIC), based on estimating a distance between the fitted and true distribution, the Bayesian Information Criterion (BIC), based on Laplace approximation of the Bayes factor between two models, and Mallows Cp, estimating the squared prediction error in a linear regression model.
Similar material can to some extent be found in Pawitan (2001), In All Likelihood, Oxford University Press. Section 13.5+13.6, as well as section 13.6 of Wasserman (2005). All of Statistics. Springer Verlag.
Last updated: Thursday, 06 March 2008 17:06. Steffen L. Lauritzen