Thomas Lumley • 7/16/2024

A Bayesian t-test, again

The article discusses Bayesian methods for two-sample mean comparisons, moving beyond traditional t-tests. It contrasts frequentist approaches (CLT, bootstrapping) with Bayesian alternatives like using weakly informative priors on summary statistics, Dirichlet process priors, and a novel 'martingale posterior' method that resamples data to generate posterior distributions for means without parametric models.

0 comments

#data analysis #Bayesian Statistics #Statistical Inference