interaction.rsf {randomSurvivalForest}R Documentation

VIMP for Single or Grouped Variables

Description

Calculate variable importance (VIMP) for a single variable or group of variables.

Usage

    interaction.rsf(object,
                  predictorNames = NULL,
                  subset = NULL,
                  joint = TRUE,
                  rough = FALSE,
                  importance = c("randomsplit", "permute", "none")[1],
                  seed = NULL,
                  do.trace = FALSE,
                  ...)

Arguments

object An object of class (rsf, grow) or (rsf, forest). Note: forest=TRUE must be used in the original rsf call.
predictorNames Character vector of variable names to be considered. This must be specified.
subset An index vector indicating which rows should be used. Default is to use all the data.
joint Should joint-VIMP or individual VIMP be calculated? See details below.
rough Logical value indicating whether fast approximation should be used. Default is FALSE.
importance Method used to compute variable importance (VIMP).
seed Seed for random number generator. Must be a negative integer (the R wrapper handles incorrectly set seed values).
do.trace Logical. Should trace output be enabled? Default is FALSE. Integer values can also be passed. A positive value causes output to be printed each do.trace iteration.
... Further arguments passed to or from other methods.

Details

Using a previously grown forest, and restricting the data to that indicated by subset, calculate the VIMP for variables listed in predictorNames. If joint=TRUE, a joint-VIMP is calculated. The joint-VIMP is the importance for the group of variables, when the group is perturbed simultaneously. If joint=FALSE, the VIMP for each variable considered separately is calculated.

Depending upon the option importance, VIMP is calculated either by random daugther assignment, by random permutation of the variable(s), or none (no perturbing).

Value

A list with the following components:

err.rate Vector of length ntree containing OOB error rates for the (unperturbed) ensemble restricted to the subsetted data.
importance Variable importance (VIMP). Either a vector or a single number depending upon the option joint.

Author(s)

Hemant Ishwaran hemant.ishwaran@gmail.com and Udaya B. Kogalur ubk2101@columbia.edu

References

H. Ishwaran (2007). Variable importance in binary regression trees and forests, Electronic J. Statist., 1:519-537.

See Also

find.interaction.

Examples

# Example of paired-VIMP.
# Veteran data.
data(veteran, package = "randomSurvivalForest") 
v.out <- rsf(Survrsf(time,status)~., veteran, ntree = 1000, forest = TRUE)
interaction.rsf(v.out, c("karno","celltype"))$importance

## Not run: 
# Individual VIMP for data restricted to events only.
# PBC data.
data(pbc, package = "randomSurvivalForest") 
rsf.out <- rsf(Survrsf(days,status)~., pbc, ntree = 1000, forest = TRUE)
o.r <- rev(order(rsf.out$importance))
VIMP <- rsf.out$importance[o.r]
VIMP.events <- rep(0, length(VIMP))
names(VIMP.events) <- names(VIMP) 
events <- which(rsf.out$cens == 1)
VIMP.events <-
 interaction.rsf(rsf.out, names(VIMP), events, joint = FALSE)$importance
VIMP.all <- as.data.frame(cbind(VIMP.events = VIMP.events, VIMP = VIMP))
print(round(VIMP.all, 3))
# PBC data again.
# Monte Carlo estimates for VIMP.
# Bootstrap estimates for VIMP.
VIMP.MC <- VIMP.BOOT <- NULL
for (k in 1:100) {
VIMP.MC <-
 cbind(VIMP.MC, interaction.rsf(rsf.out, names(VIMP), joint = FALSE)$importance)
VIMP.BOOT <-
 cbind(VIMP.BOOT, interaction.rsf(rsf.out, names(VIMP),
      subset = sample(1:dim(pbc)[1], replace = TRUE), joint = FALSE)$importance)
}
VIMP.MC <- as.data.frame(cbind(VIMP.mean = apply(VIMP.MC, 1, mean),
                               VIMP.sd = apply(VIMP.MC, 1, sd)))
VIMP.BOOT <- as.data.frame(cbind(VIMP.mean = apply(VIMP.BOOT, 1, mean),
                               VIMP.sd = apply(VIMP.BOOT, 1, sd)))
rownames(VIMP.MC) <- rownames(VIMP.BOOT) <- names(VIMP)
print(round(VIMP.MC, 3))
print(round(VIMP.BOOT, 3))
## End(Not run)

[Package randomSurvivalForest version 3.2.3 Index]