Demographically Stratified Templates for MAGeTbrain

While the MAGeTbrain algorithm has been shown to be insensitive to the choice of templates, referees will typically expect template selection to be stratified across age/sex and other factors.

The following is example code that takes a dataset with ID, Age and Sex, and calculates an age-stratitifed, sex balanced template set from the original data. If you also need to split based on some group, this step would be done similar to the Age splitting.

#filtered.data is only valid subjects (QC == 1)

#Generate Templates for MAGET - Split Male/Female, Break into Bins by Age, Pick one Subject from Each Age Bin
#Count number of women and men
nummen = round(21 * dim(subset(filtered.data, M.F == "M"))[1] / dim(filtered.data)[1])
numwomen = round(21 * dim(subset(filtered.data, M.F == "F"))[1] / dim(filtered.data)[1])

#Sort subjects by age
sorted.filtered.data = filtered.data[order(filtered.data$Age),]

#Calculate bins in Age for Men and women
fembreaks=seq(min(subset(sorted.filtered.data, M.F=="F")$Age), max(subset(sorted.filtered.data, M.F=="F")$Age), length.out=numwomen+1)
manbreaks=seq(min(subset(sorted.filtered.data, M.F=="M")$Age), max(subset(sorted.filtered.data, M.F=="M")$Age), length.out=nummen+1)

#Group subjects into bins based on bin calculation above
femgroups = split(subset(sorted.filtered.data, M.F=="F"), cut(subset(sorted.filtered.data, M.F=="F")$Age, fembreaks))
mangroups = split(subset(sorted.filtered.data, M.F=="M"), cut(subset(sorted.filtered.data, M.F=="M")$Age, manbreaks))

#Create an empty list
template = c()
#Add the first subject to the list
template = append(template, as.character(mangroups[[1]]$ID[1]))
#Loop over the binned subjects and choose the middle member, add to list
for (i in seq(2,nummen-1))
  template = append(template, as.character(mangroups[[i]]$ID[ceiling(length(mangroups[[i]]$ID)/2)]))
#Add the last member
template = append(template,as.character(tail(mangroups[[nummen]]$ID,1)))

#Do the same as above for the women
template = append(template,as.character(femgroups[[1]]$ID[1]))
for (i in seq(2,numwomen-1)) 
  template = append(template,as.character(femgroups[[i]]$ID[ceiling(length(femgroups[[i]]$ID)/2)]))
template = append(template,as.character(tail(femgroups[[numwomen]]$ID,1)))

#Write out ID names to a file
write.table(template, file="templates.txt", quote=FALSE, row.names=FALSE, col.names=FALSE )