User talk:Wawaph

Akaike information criterion
Welcome to Wikipedia. Following the WP:BRD cycle, I have reverted your contribution to Akaike information criterion. I disagree with your characterization of McElreath and Taddy being wrong. Not just because one can easily find other sources confirming this definition (e.g., Cao & Soofi), but because the notion of AIC being an estimator of "in-sample" error doesn't make sense. The "prediction error" of the model within sample does not need to be estimated, it can be readily calculated. AIC sure is a function of in-sample statistics, but it aims to predict out-of-sample. --bender235 (talk) 00:34, 2 November 2020 (UTC)