Hjort, Nils Lid; Claeskens, Gerda Frequentist model average estimators. (English) Zbl 1047.62003 J. Am. Stat. Assoc. 98, No. 464, 879-899 (2003). Summary: The traditional use of model selection methods in practice is to proceed as if the final selected model had been chosen in advance, without acknowledging the additional uncertainty introduced by model selection. This often means underreporting of variability and too optimistic confidence intervals. We build a general large-sample likelihood apparatus in which limiting distributions and risk properties of estimators post-selection as well as of model average estimators are precisely described, also explicitly taking modeling bias into account. This allows a drastic reduction in complexity, as competing model averaging schemes may be developed, discussed, and compared inside a statistical prototype experiment where only a few crucial quantities matter. In particular, we offer a frequentist view on Bayesian model averaging methods and give a link to generalized ridge estimators. Our work also leads to new model selection criteria. The methods are illustrated with real data applications. Cited in 6 ReviewsCited in 201 Documents MSC: 62A01 Foundations and philosophical topics in statistics 62F12 Asymptotic properties of parametric estimators 62F25 Parametric tolerance and confidence regions 62C12 Empirical decision procedures; empirical Bayes procedures Keywords:bias and variance balance; growing models; likelihood inference; model average estimator; model information criteria; moderate misspecification; logistic regression; empirical Bayes PDFBibTeX XMLCite \textit{N. L. Hjort} and \textit{G. Claeskens}, J. Am. Stat. Assoc. 98, No. 464, 879--899 (2003; Zbl 1047.62003) Full Text: DOI Link