The reciprocal Bayesian bridge for left-censored data

Abstract

We propose two Bayesian methods for regularized left censored regression: the reciprocal Bayesian bridge and the reciprocal Bayesian adaptive bridge. Gibbs samplers are derived based on the reciprocal Bayesian bridge prior which can be written as a scale mixture of inverse uniform distribution. The proposed approaches are then illustrated via five simulated studies and a real data example. Compared with some existing methods, our methods have improved variable selection and estimation performance in both simulations and the real data example.

Keywords:

• Bayesian inference
• Bayesian model selection
• Gibbs sampler
• Reciprocal bridge
• Regularization
• Tobit regression