Hi!

I am using NanoStringDiff to analyse Nanostring mRNA data.

I use the function `PlotsPositiveHousekeeping`

to check the variation of housekeeping genes (hkg). According to the code of this function (see below), counts for hkg are scaled using a size factor calculated from the coefficients of a generalized linear model with positive and negative controls counts as response variable and the concentration of the positive controls as predictor.

I am wondering why square-root is applied to the size factor when hkg counts are standardized ?

```
coeff = c # real coefficients
c = c/mean(c) # size factors
for (i in 1:nrow(housekeeping)){
housekeeping[i,]<-(housekeeping[i,]-mean.neg)/c^(1/2)
}
```

Complete code of `PlotsPositiveHousekeeping`

from https://github.com/tingtina/NanoStringDiff/blob/master/R/PlotsPositiveHousekeeping.R

```
PlotsPositiveHousekeeping<- function(path=path, header=TRUE){
# ------------------------------------ #
# path: points out the directory in which your csv.file is located #
# header: a logical value(TRUE or FALSE) indicating whether the file contains the names of the variables as its first line. #
# designs: data frame for pheno type data storage #
# ------------------------------------ #
data = read.table(path, header = header, sep = ",")
if (is.null(data)) {
stop("There is no counts data")
}
selectcol = !(names(data) %in% c("Code.Class", "Name", "Accession"))
## remove NA from data set
counts = data[,selectcol]
counts = as.matrix(counts)
id = which(is.na(rowSums(counts)))
if (length(id) > 0) {
data = data[-id, ]
}
code.class = tolower(data[, c("Code.Class")])
name = data[, c("Name")]
accession = data[, c("Accession")]
counts = data[,selectcol]
counts = as.matrix(counts)
rownames(counts) = name
pos.id = grep("positive", code.class, fixed = TRUE)
neg.id = grep("negative", code.class, fixed = TRUE)
house.id = grep("housekeeping", code.class, fixed = TRUE)
spikein.id = grep("spikein", code.class, fixed = TRUE)
positive = counts[pos.id, ]
negative = counts[neg.id, ]
housekeeping = counts[house.id, ]
house.adjust = counts[house.id, ]
mean.neg<-colMeans(negative)
pos = positive
pos[pos <= 0] = 1
neg = negative
x = c(128, 32, 8, 2, 0.5, 0.125, rep(0, nrow(neg)))
Y = rbind(pos, neg)
n = ncol(Y)
c = rep(0, n)
for (i in 1:n) {
model = glm(Y[, i] ~ x, family = poisson(link = identity))
c[i] = model$coeff[2]
}
coeff = c # real coefficients
c = c/mean(c) # size factors
for (i in 1:nrow(housekeeping)){
housekeeping[i,]<-(housekeeping[i,]-mean.neg)/c^(1/2)
}
nhouse<-nrow(housekeeping)
varhouse<-rowVars(housekeeping)^(1/2)/rowMeans(housekeeping)
con=c(128,32,8,2,0.5,0.125)
par(mfrow=c(1,2),mar=c(5,6,2,1)+0.1, oma=c(2,2,2,2)+0.1)
plot(con,rowMeans(positive),xlab="Concentration",ylab="Positive Counts",
col = "blue")
abline(lm(rowMeans(positive) ~ con))
plot(c(1:nhouse),varhouse,xaxt='n',xlab="",ylab="Coefficient of Variation",
col = "blue", xlim =c(1,nhouse+1), ylim =c(min(varhouse)*0.9,max(varhouse)*1.1))
title(xlab="Housekeeping Genes", line=1.5)
textxy(c(1:nhouse), varhouse, labs=rownames(housekeeping), cex = 1, m = c(0, 0), offset = 0.65)
} # end function #
```