Sometimes, 'dbarts' entirely drops a variable from the model without even including it in the variable splits with "0" splits - both can happen in the same model, if dimensionality is high. This causes issues with varimp() and then some downstream issues.
I started writing a solution for varimp() but it doesn't work yet, and a lot of things are broken downstream because of it.
Sometimes, 'dbarts' entirely drops a variable from the model without even including it in the variable splits with "0" splits - both can happen in the same model, if dimensionality is high. This causes issues with varimp() and then some downstream issues.
I started writing a solution for varimp() but it doesn't work yet, and a lot of things are broken downstream because of it.