CMStatistics 2022: Start Registration
View Submission - CMStatistics
Title: Type I Tobit Bayesian additive regression trees for censored outcome regression Authors:  Eoghan O Neill - Erasmus University Rotterdam (Netherlands) [presenting]
Abstract: Type 1 Tobit Bayesian Additive Regression Trees (TOBART-1) are introduced. Simulation results and applications to real datasets demonstrate that TOBART-1 produces more accurate predictions than competing methods. TOBART-1 provides posterior probabilities of censoring, posterior intervals for the conditional expectation, and estimates of heterogeneous treatment effects. TOBART-1 is also combined with a Dirichlet Process mixture of normal distributions to provide a fully nonparametric censored outcome regression method (TOBART-1-NP).