Using Bayesian hierarchical models for controlled post hoc subgroup analysis of clinical trials: application to smoking cessation treatment in American Indians and Alaska Natives.