var.imp {gbev}R Documentation

Variable importance.

Description

Computes and plots the variable importance of variables used in model.

Usage

var.imp(object,nboosts=NULL,std=FALSE,plt=TRUE)

Arguments

object A fitted object of type gbev.object.
nboosts Number of boosting iterations, if NULL set to maximum number used.
std If FALSE then the reduction in mean squared error due to covariates is given, if TRUE then this reduction is standardized with respect to variance of response.
plt If TRUE the importances are plotted.

Details

Variable importance is defined as the reduction in mean squared error due to the different covariates. For a given covariate, this is computed by idenifying all nodes split using that covariate and adding together the reductions in mean squared error due to splitting these nodes.


[Package gbev version 0.1.1 Index]