**0**wrote:

I'm using glmLRT() from edgeR to analyze some RNA-seq data. My experimental design has one Treatment factor with three classes (A, B and C), one continuous Age covariate and one Batch factor with 19 levels.

`> colnames(design)`

[1] "(Intercept)" "ClassB" "ClassC" "Age"
[5] "Batch1" "Batch2" "Batch3" "Batch4"
[9] "Batch5" "Batch6" "Batch7" "Batch8"
[13] "Batch9" "Batch10" "Batch11" "Batch12"
[17] "Batch13" "Batch14" "Batch15" "Batch16"
[21] "Batch17" "Batch18" "Batch19”

This is an ANCOVA design, so when I make my comparisons between the three Treatment classes (A, B, C), I have been taught that I need to pick a specific covariate value to make the contrasts. Typically you might pick Age = 0, Age = min(Age), Age = mean(Age or Age = max(Age) for your contrasts. I've been estimating the contrasts for the mean Age. Let's suppose the mean age is 7.3 years for this example. I'm using the code

`> lrt1 = glmLRT(fit,contrast=c(0,1,0,mean(Age),rep(0,19)))`

... to compute my contrasts. The code runs and the output makes sense, but my collaborators would like some assurance that glmLRT() is computing the correct contrast. Unfortunately, there are no explanations of covariate effects and contrasts in the edgeR manuals or tutorials, and I can't find any information on Google either. Can glmLRT compute contrasts adjusted for a continuous covariate like this? Or am I computing nonsense here?