Skip to contents

flexBART 2.0.5

Major changes

  • Refactor internal function to detect nesting structure, which previously threw an error if it encountered a categorical variable that was incompletely observed (i.e., not every level appeared in the supplied data)
  • Fixes error in how heteroskedastic BART saves samples of the residual standard deviation function with test data.

flexBART 2.0.0

Major changes

  • Add support for varying coefficient models (i.e., multipe additve ensembles)
  • Remove unnecessary passes over the full training data during backfitting and computing conditional posterior of leaf nodes
  • Introduce pre-processing functions
  • Add support for nested categorical features
  • Introduce formula interface

flexBART 1.0.0

Major changes

  • Initial package