Bootstrap Methods and Their Application

sive models further, It gives additional illustrations of the performance of selection criteria in small samples (e.g., with respect to several varying characteristics that include sample size and parameter structure), in large samples (e.g., with respect to the degree of overfitting allowed), and on real data. This book looks at many selection criteria for a wide range of models, from traditional models to more recently developed classes of models. Currently, however, there appear to be few conclusive rules that give the best performing selection criteria for a given setting for this subject. As a result, the greatest value of this book appears to be in its derivation/description of the various selection criteria and the presentation of approaches for evaluating selection-criterion performance. The book also contains several instances in which key terms are not defined explicitly and no proofs or arguments are given for key claims. One message that summarizes several results given in this book is that certain selection criteria developed for traditional regression and time series models (e.g., Al&), when naively applied to certain nonnormal settings, appear to perform at least as well as selection criteria specifically designed for those settings.