Some Model Selection Criteria based on Incomplete or Complete data

This book present some extensions of model selection criteria based on incomplete or complete data. First, we have derived a corrected version of the KIC criterion, based on the Kullback's symmetric divergence for multiple and multivariate regression models and for univariate and multivariate autoregressive models. Its signal-to-noise ratios is pointed out in each case. In the presence of missing data, we derived and investigated a variant of KIC criterion. We examine the performance of the new criterion relative to other well known criteria in a large simulation study. Also, we present a variants of a Schwarz information criterion for model selection in the settings where the observed-data is incomplete. The performance of these criteria, relative to other well known criteria, is examined in a large simulation study. Finally, we study the asymptotic property and the performance of the repeated half sampling (RHS) criterion. Two cases are distinguish for the true model, either the candidate family of models does include the true model or does not include it. The performance of RHS criterion is compared with other criteria in a simulation study.