predict.flexBART: Predicting new observations with previously fitted flexBART...

In flexBART: A More Flexible BART Model

Predicting new observations with previously fitted flexBART model

Description

predict.flexBART() can take the output of flexBART and use it to make predictions at new inputs.

Usage

## S3 method for class 'flexBART'
predict(object, newdata, ...)

Arguments

object	object of class inheriting from “flexBART”.
newdata	Data frame in which to look for variables with which to predict. Cannot be omitted.
...	Additional optional arguments governing whether to print progress (i.e., verbose and print_every).

Details

Make predictions at new inputs based on the output of flexBART. When the training and/or testing dataset is large, it is recommended to run flexBART with the arguments save_samples = FALSE and save_trees = TRUE. Then, to access posterior samples of the function evaluations, pass the fitted “flexBART” object to predict.flexBART() along with the appropriate newdata argument. For probit or logistic BART, the function outputs draws of the fitted probabilities.

Value

When there is only one ensemble, a matrix containing posterior samples of the regression function evaluated at the supplied inputs. For models with multiple ensembles, a list with three elements: (i) a matrix containing posterior samples of the regression function evaluation and (ii) an array containing evaluations of the identified slopes; and (iii) an array containing evaluations of all slopes on the standardized scale.

flexBART documentation built on Oct. 2, 2026, 1:07 a.m.