poolVIM {poolVIM}

after the Actual Impurity Reduction Importance is computed with a Random Forest, pvalues from different probes or SNPs belonging to the same gene can be aggregated in order to obtain a single pvalue for that gene. Correlation between probes can also be taken into account.

Description

after the Actual Impurity Reduction Importance is computed with a Random Forest, pvalues from different probes or SNPs belonging to the same gene can be aggregated in order to obtain a single pvalue for that gene. Correlation between probes can also be taken into account.

Usage

poolVIM(rf, genenames, x, method = "Tippett", adjust)

Arguments

rf

a ranger object with "importance="impurity_corrected"

genenames

a vector of the name of the gene to which each probe or SNP belongs, it has to be of size dim(x)[1]

x

design matrix used by the random forest

method

one of Tippett, Fisher, Kost, EBM

adjust

"no" / "yes" depending if correlation has to be taken into account

Examples

n <- 250
x=replicate(50, runif(n))
dat <- data.frame(y = factor(rbinom(n, 1, .5)), x)
library(ranger)
rf <- ranger(y ~ ., dat, importance = "impurity_corrected",num.trees=100)
genenames=colnames(x)=rep(c("G1","G2"),50/2)
poolVIM(rf,genenames,x,method="Fisher",adjust="no")