Build multilevel Bayesian models, run prior and posterior predictive checks, and compare models using predictive performance. Revise models through a transparent cycle of criticism and improvement.
