Talk:RankProduct.R
From Organic Design wiki
Error in RankProd package, specifically related to the RP and topGene function. The help(RP) documentation provides the following information about NA handling; na.rm: if 'FALSE' (default), the NA value will not be used in computing rank. If 'TRUE', the missing values will be replaced by the gene-wise mean of the non-missing values. Gene with all values missing will be assigned "NA" So, in the case where an entire row of observations for a gene are NAs the rank product should be assigned NAs. Generating a 10 by 4 matrix from a normal distribution with two rows of missing values; library(RankProd) set.seed(1) nrows <- 10 ncols <- 4 X <- round(matrix(rnorm(nrows * ncols), nc=ncols),2) X[2:3,] <- NA X RP.out <- RP(X, cl=rep(1,ncols), na.rm=TRUE) RP.out$RPrank RP.out$RPrank[2:3,] # Small bug here subset(RP.out$Orirank, 2:3) RP.out$Orirank[["class1 > class 2"]] # (additional space in list name) RankProd1(X, logged=TRUE, num.class=1, rev.sorting=TRUE)$rank.all RP.out$Orirank[["class1 < class2"]] RankProd1(X, logged=TRUE, num.class=1, rev.sorting=FALSE)$rank.all RP.out$pval tG <- topGene(RP.out, cutoff = 0.1, method="pval", logged=FALSE, logbase=2, gene.names=rownames(X)) # List is wrong tG[[1]] # List ok tG[[2]] sessionInfo()