A/B testing and randomized experiments playbook: hypothesis testing, confidence intervals, sample size, sequential testing (mSPRT), and multi-armed bandits

A comparison between Bayesian and heuristic machine-learning models that learn probability information from experience under uncertainty.

Production-style A/B testing with binomial GLMs (logit/probit): covariate adjustment, marginal ATE/risks, cluster-robust SEs, and Brier-score calibration.