website

Statistical Modeling, Causal Inference, and Social Science

by Andrew Gelman

Andrew Gelman and his co-bloggers post almost daily on statistics as it's actually practised: forking paths, overconfident p-values, shaky survey adjustments, and how weak studies get published and then defended. The commenters are mostly working statisticians, and a year of reading it is a better education in the replication crisis than most courses.