Bayesian Analysis of Randomized Experiments with Non-Compliance
Functions for Bayesian analysis of data from randomized experiments with non-compliance. The functions are based on the models described in Imbens and Rubin (1997) <doi:10.1214/aos/1034276631>. Currently only two types of outcome models are supported: binary outcomes and normally distributed outcomes. Models can be fit with and without the exclusion restriction and/or the strong access monotonicity assumption. Models are fit using the data augmentation algorithm as described in Tanner and Wong (1987) <doi:10.2307/2289457>.