Model choice

Proceedings of the 4th Franco-Belgian Meeting of Statisticians, november 1983
Sous la direction de
M. Mouchart
L. Simar

This volume contains a collection of papers which were presented at the fourth Franco-Belgian Meeting of Statisticians, held November 24 and 25, 1983 in Belgium. The articles which are gathered together here present different approaches to the major theme of the meeting : model choice. Insofar as the choice of a model does not constitute a well-defined statistical problem but rather has to do with the point of view one holds about the given information, it is not at all surprising that such a variety of ideas exists on this topic. Model choice can appear to be a hypothesis testing problem, nested or not, whose asymptotical theory demands a theoretical detour by way of the contiguity notion. It can also take the form of an estimation of a discrete parameter (from a Bayesian view or not). An example would be in discriminant analysis or like a non-parametric test problem. While working on chronological series, it is very common to choose between various models depending upon the given information. This volume examines both theoretically, and through applications, the subject of model choice.