trainModelVarSelSD: Extract cross-validated important variables

In environmentalinformatics-marburg/gpm: Geospatial predictive modeling using parameterized and unparameterized models.

Description

This function approaches the identification of important variables from rfe more conservatively than caret. It uses the standard deviation (or standard error) of the cross-validated error metric to identify important variables.

Usage

trainModelVarSelSD(model, metric = model$metric, maximize = FALSE,
  sderror = TRUE)

Arguments

model	a rfe model. See rfe
metric	the metric to be used. Note this needs to be the metric used to calculate the rfe model
maximize	logical: Is a higher value of the metric favourable (e.g metric = Rsquared) or not (e.g metric = RMSE). maximize=TRUE is determined automatically as long as metric is either Rsquared, ROC, Accuracy. maximize =FALSE is used for all other metrics. Set this manually if you use an other metric where higher values are favourable.
sderror	If TRUE then standard error is calculated. If FALSE then standard deviations are used

Value

a character vector of the variable names

Examples

# Not run

environmentalinformatics-marburg/gpm documentation built on July 11, 2020, 11:12 a.m.