R/runWilcox.R

In montilab/K2Taxonomer: Creating annotated submodules of high-throughput data through recursive partitioning

## Create data.frame of factor variable
.runWilcox <- function(csDF, alternative="two.sided") {

    ## Run test
    if (sum(!is.na(csDF$value[csDF$cat])) > 2 &
        sum(!is.na(csDF$value[!csDF$cat])) > 2) {
        wilcoxOut <- suppressWarnings(wilcox.test(csDF$value[csDF$cat],
            csDF$value[!csDF$cat], alternative=alternative))
        pval <- wilcoxOut$p.value
        stat <- wilcoxOut$statistic
    } else {
        pval <- NA
        stat <- NA
    }

    ## Get means of each group and diff
    obsMean <- mean(csDF$value[csDF$cat], na.rm=TRUE)
    altMean <- mean(csDF$value[!csDF$cat], na.rm=TRUE)
    diffMean <- obsMean - altMean

    ## Get type of test
    test <- c("one-sided Wilcox RS Test",
        "two-sided Wilcox RS Test")[as.numeric(alternative == "two.sided") + 1]

    ## Generate results
    data.frame(pval=pval, stat=stat, df=NA, obsMean=obsMean,
        altMean=altMean, diffMean=diffMean, nhits=NA, ncase=NA,
        nalt=NA, ndrawn=NA, hits=NA, test="1-sided Wilcox Test",
        stringsAsFactors=FALSE)
}