Revision b7c0afd9fdb886458ef54c202368a21551f6b1b6 authored by Max Kuhn on 11 June 2008, 14:44:07 UTC, committed by cran-robot on 11 June 2008, 14:44:07 UTC
1 parent 6a94bf7
varImp.Rd
\name{varImp}
\alias{varImp}
\alias{varImp.train}
\alias{varImp.earth}
\alias{varImp.rpart}
\alias{varImp.randomForest}
\alias{varImp.gbm}
\alias{varImp.regbagg}
\alias{varImp.classbagg}
\alias{varImp.pamrtrained}
\alias{varImp.lm}
\alias{varImp.mvr}
\alias{varImp.bagEarth}
\alias{varImp.RandomForest}
\title{Calculation of variable importance for regression and classification models}
\description{
A generic method for calculating variable importance for objects produced by
\code{train} and method specific methods
}
\usage{
\method{varImp}{train}(object, useModel = TRUE, nonpara = TRUE, scale = TRUE, ...)
\method{varImp}{earth}(object, value = "gcv", ...)
\method{varImp}{rpart}(object, ...)
\method{varImp}{randomForest}(object, ...)
\method{varImp}{gbm}(object, numTrees, ...)
\method{varImp}{classbagg}(object, ...)
\method{varImp}{regbagg}(object, ...)
\method{varImp}{pamrtrained}(object, threshold, data, ...)
\method{varImp}{lm}(object, ...)
\method{varImp}{mvr}(object, ...)
\method{varImp}{bagEarth}(object, ...)
\method{varImp}{RandomForest}(object, normalize = TRUE, ...)
}
\arguments{
\item{object}{an object corresponding to a fitted model}
\item{useModel}{use a model based technique for measuring variable importance?
This is only used for some models (lm, pls, rf, rpart, gbm, pam and mars)}
\item{nonpara}{should nonparametric methods be used to assess the relationship
between the features and response (only used with \code{useModel = FALSE} and
only passed to \code{filterVarImp}).}
\item{scale}{should the importances be scaled to 0 and 100?}
\item{\dots}{parameters to pass to the specific \code{varImp} methods}
\item{numTrees}{the number of iterations (trees) to use in a boosted tree model}
\item{threshold}{the shrinkage threshold (\code{pamr} models only)}
\item{data}{the training set predictors (\code{pamr} models only)}
\item{value}{the statistic that will be used to calculate importance:
either \code{gcv}, \code{nsubsets}, or \code{rss}}
\item{normalize}{a logical; should the OOB mean importance values be divided
by their standard deviations?}
}
\value{
A data frame with class \code{c("varImp.train", "data.frame")} for
\code{varImp.train} or a matrix for other models.
}
\details{
For models that do not have corresponding \code{varImp} methods, see
\code{filerVarImp}.
Otherwise:
\item Linear Models: the absolute value of the t--statistic
for each model parameter is used.
\item Random Forest: \code{varImp.randomForest} and
\code{varImp.RandomForest} are wrappers around the importance functions from the
\pkg{randomForest} and \pkg{party} packages, respectively.
\item Partial Least Squares: the variable importance measure here is based on
weighted sums of the absolute regression coefficients. The weights are a function of
the reduction of the sums of squares across the number of PLS components and are
computed separately for each outcome. Therefore, the contribution of the coefficients
are weighted proportionally to the reduction in the sums of squares.
\item Recursive Partitioning: The reduction in the loss function
(e.g. mean squared error) attributed to each variable at each split is
tabulated and the sum is returned. Also, since there may be candidate variables
that are important but are not used in a split, the top competing variables are
also tabulated at each split. This can be turned off using the \code{maxcompete}
argument in \code{rpart.control}. This method does not currently provide
class--specific measures of importance when the response is a factor.
\item Bagged Trees: The same methodology as a single tree is applied to
all bootstrapped trees and the total importance is returned
\item Boosted Trees: \code{varImp.gbm} is a wrapper around the function from that package (see the \pkg{gbm} package vignette)
\item Multivariate Adaptive Regression Splines: MARS models
include a backwards elimination feature selection routine that
looks at reductions in the generalized cross-validation (GCV)
estimate of error. The \code{varImp} function tracks the changes in
model statistics, such as the GCV, for each predictor and
accumulates the reduction in the statistic when each
predictor's feature is added to the model. This total reduction
is used as the variable importance measure. If a predictor was
never used in any of the MARS basis functions in the final model
(after pruning), it has an importance
value of zero. Prior to June 2008, the package used an internal function
for these calculations. Currently, the \code{varImp} is a wrapper to
the \code{\link[earth]{evimp}} function in the \code{earth} package. There are three statistics that can be used to
estimate variable importance in MARS models. Using
\code{varImp(object, value = "gcv")} tracks the reduction in the
generalized cross-validation statistic as terms are added.
However, there are some cases when terms are retained
in the model that result in an increase in GCV. Negative variable
importance values for MARS are set to zero.
Alternatively, using
\code{varImp(object, value = "rss")} monitors the change in the
residual sums of squares (RSS) as terms are added, which will
never be negative.
Also, the option \code{varImp(object, value =" nsubsets")}, which
counts the number of subsets where the variable is used (in the final,
pruned model).
\item Nearest shrunken centroids: The difference between the class centroids and the overall centroid is used to measure the variable influence (see \code{pamr.predict}). The larger the difference between the class centroid and the overall center of the data, the larger the separation between the classes. The training set predictions must be supplied when an object of class \code{pamrtrained} is given to \code{varImp}.
}
\author{Max Kuhn}
\keyword{ models }
Computing file changes ...