Bayesian mismeasurement t-models for censored responses