Efficient and reliable software is critical to the successful usage of the BART methodology. We believe that this package fills that need. The BART engine provided here is written in C++ for reasons of speed and maintainability; the source code is easy to read and modify while still being relatively efficient computationally. This package brings together R functions for a variety of outcomes: continuous, dichotomous, categorical and time-to-event with right censoring.
For continuous outcomes, use the wbart/mc.wbart function(s).
For dichotomous outcomes, use the pbart/mc.pbart or
lbart/mc.lbart function(s).
For categorical outcomes, use the mbart/mc.mbart function(s).
For survival outcomes, use the surv.bart/mc.surv.bart function(s).
For competing risk outcomes, use the crisk.bart/mc.crisk.bart function(s).
For recurrent event outcomes, use the recur.bart/mc.recur.bart function(s).
For convenience, these functions return the trees (so that you can make
predictions without fitting the model simultaneously) via the predict function.