Dan Simpson • 12/8/2021

Why won’t you cheat with me? (Repost)

This article explores the use of scaled priors in Bayesian statistics, specifically focusing on priors that promote sparsity in high-dimensional models. It discusses the computational limitations of spike-and-slab models and advocates for scale-mixture of normal priors (local-global priors) as a practical alternative for achieving approximate sparsity, followed by decision rules for exact sparsity.

0 comments

#Statistical Modeling #Bayesian Statistics #Experimental Design