Statistical Aspects of Model Selection

3267 19 3308 3 6575 document 3308 WP-89-077.pdf application/pdf 71093c275d5aa837e2d50d3d40700aad MD5 980370 2016-01-15 01:59:36 https://pure.iiasa.ac.at/id/eprint/3267/1/WP-89-077.pdf 3267 1 1 text en public

WP-89-077.pdf

22 2b444ac54f57f2c017672690812aafb83fe6244a 17348 2 29078 document 17348 lightbox.jpg image/png b03554f4745bfaf40ad4176eabca2d44 MD5 11174 2016-01-15 12:18:26 https://pure.iiasa.ac.at/id/eprint/3267/2/lightbox.jpg 3267 2 2 image/png other Thumbnails conversion from text to thumbnail_lightbox en public

lightbox.jpg

preview.jpg

medium.jpg

small.jpg

indexcodes.txt

WP-89-077.pdf

http://eprints.org/relation/isVersionOf https://pure.iiasa.ac.at/id/document/3308 http://eprints.org/relation/isVolatileVersionOf https://pure.iiasa.ac.at/id/document/3308 http://eprints.org/relation/isCoversheetVersionOf https://pure.iiasa.ac.at/id/document/3308 archive 351 disk0/00/00/32/67 2016-01-15 01:59:36 2021-08-27 17:13:33 2016-01-15 01:59:36 monograph show 2 Shibata R. Statistical Aspects of Model Selection pub iis_met iis_mod prog_sds Various aspects of statistical model selection are discussed from the view point of a statistician. Our concern here is about selection procedures based on the Kullback Leibler information number. Derivation of AIC (Akaike's Information Criterion) is given. As a result a natural extension of AIC, called TIC (Takeuchi's Information Criterion) follows. It is shown that the TIC is asymptotically equivalent to Cross Validation in a general context, although AIC is asymptotically equivalent only for the case of independent identically distributed observations. Next, the maximum penalized likelihood estimate is considered in place of the maximum likelihood estimate as an estimation of parameters after a model is selected. Then the weight of penalty is also the one to be selected. We will show that, starting from the same Kullback-Leibler information number, a useful criterion RIC (Regularization Information Criterion) is derived to select both the model and the weight of penalty. This criterion is in fact an extension of TIC as well, as of AIC. Comparison of various criteria, including consistency and efficiency is summarized in Section 5. Applications of such criteria to time series models are given in the last section. 1989-10 published WP-89-077 WP-89-077 10 public working_paper IIASA, Laxenburg, Austria 34 FALSE info:eu-repo/semantics/book