Richard McElreath 🐈‍⬛

New paper "To select or not to select: predictively consistent priors instead of model selection" with Anna Elisabeth Riha, Leevi Lindgren, @davidkohns.bsky.social, @paulbuerkner.com arxiv.org/abs/2606.22850 Model selection is not a substitute for building good models in the f…

To select or not to select: predictively consistent priors instead of model selection arxiv.org

AI Weekly's analysis →

• The paper claims flexible models with predictively consistent priors typically match or outperform selected simpler models in out-of-sample predictive performance.
• When model selection does help, the authors argue it signals problematic prior specifications rather than a genuine modeling advantage.
• The framework is tested across linear and logistic regression, variable selection, and nonlinear modeling, in a 34 page paper with 36 supplementary pages.

Read full analysis →

View on Bluesky →