Bayesian analysis of null intercept errors-in-variables regression for pretest/post-test data

This article discusses a Bayesian analysis of repeated measures pretest/post-test data under null intercepts errors-in-variables regression models. For illustration we consider an example in the field of dentistry involving the comparison of two types of toothbrushes with respect to the efficacy in removing dental plaque. The proposed Bayesian approach accommodates the correlated measurements and incorporates the restriction that the slopes must lie in the [0,1] interval, a feature not considered in the analysis conducted by Singer & Andrade (1997). The observed values of the (repeated) response and explanatory variables are supposed to follow a Multivariate Student- t distribution. A Gibbs sampler is used to perform the computations.