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