Enable group-level weight specification in flm_def constrast construction

[lianos]

You can combine multiple groups in the numer or denom when constructing a linear model.

Currently it gives each group within the numer/denom equal weight, but you may want to adjust that.

Using the example tcga dataset, we can compare stage1 and stage2 vs stage3 and stage4 tumors like so:

library(FacileAnalysis)
library(FacileData)

efds <- exampleFacileDataSet()
es <- filter_samples(efds, indication == "BLCA", sample_type == "tumor")

flm.equal <- flm_def(es, "stage", numer = c("I", "II"), denom = c("III", "IV"))

That would construct a contrast vector that compares the average of stage I and II vs the average of stage III and IV, like so:

   I   II  III   IV 
 0.5  0.5 -0.5 -0.5 

But what if you wanted something like this?

    I    II  III   IV 
 0.75  0.25 -0.5 -0.5 

Potential Syntax

Option 1

The names of the numer/denom character vectors could be the group names, and the values could be the fractional weights:

flm.weighted <- flm_def(es, numer = c("I" = 0.75, "II" = 0.25), denom = c("III", "IV"))

This would require us to check the numer/denom vectors to see if they were numeric or character first, and then pull the names of the groups and their weights from the appropriate.

Option 2

Or vice versa? This way may require less (code) hacking, but I'm not the biggest fan of havig numbers for names() of a vector

flm.weighted <- flm_def(es, numer = c("0.75" = "I", "0.25" = "II"), denom = c("III", "IV"))

Option 3

You could imagine having a separate parameter for group weights, but ...