variable_importance: variable_importance

In AndrewLawrence/dCVnet: doubly cross-validated elastic-net regularised generalised linear models

variable_importance

Description

Variable importance for dCVnet/glmnet models does not require permutation methods, because coefficients are directly interpretable.

Usage

variable_importance(x, scale = FALSE, percentage = FALSE)

Arguments

x	a dCVnet object
scale	Boolean. Should the return values be scaled so the most important value is 1?
percentage	Boolean. Should the return values be scaled so the most important value is 100?

Details

This VI function follows caret's example (see \code{\link[caret]{varImp}}
function) and simply returns the absolute values of the coefficients.

As variable importance is inferential this is done for the tuned
"production" model rather than the cross-validated outer-loop.

Value

a data.frame of variable names "Predictor" and variable importance "varImp".

See Also

varImp

AndrewLawrence/dCVnet documentation built on Sept. 24, 2024, 5:24 a.m.