You can use an R 'transform' function to transform the data and pass that function to the RevoScaleR 'rxDataStepXdf()' function. You can then use the newly created, subset .xdf file with other RevoScaleR functions. Below is a sample R script that creates a new .xdf file by randomly sampling a larger .xdf file using the hidden row selection variable available in 'transformFunc'.
# Create a transformFunc that selects 25% of the data at random
set.seed(13) xform <- function(data) { data$.rxRowSelection<-as.logical(rbinom(length(data[[1]]),1,.25)) return(data) } rxDataStepXdf(inFile=inFile, outFile="sampledData.xdf", transformFunc=xform, overwrite=TRUE) # check that subsetting was done and the row selection variable is not kept in the data set. rxGetInfoXdf(inFile) rxGetInfoXdf("sampledData.xdf")