variableImportance: Variable importance based on MMD

In drf: Distributional Random Forests

Description

compute an mmd-based variable importance for the drf fit.

Usage

variableImportance(
  object,
  h = NULL,
  response.scaling = TRUE,
  type = "difference"
)

Arguments

object	an S3 object of class drf.
h	the bandwidth parameter, default to NULL using then the median heuristic.
response.scaling	a boolean value indicating if the responses should be scaled globally beforehand.
type	the type of importance, could be either "raw", the plain MMD values, "relative", the ratios to the observed MMD or "difference", the excess to the observed MMD

Value

a vector of variable importance values.

drf documentation built on March 29, 2021, 5:07 p.m.