Hi, I'm running differential transcript and exon usage analyses using DEXSeq and have a quick question about when I should subset the data.

For the purposes of my experiment, we have multiple nested groups of contrasts that we are interested in. Unlike DESeq2 we can't just fit a more complex model, do multiple group comparisons and pull out individual contrasts and have to subset the data beforehand which is absolutely fine. We are confident that we have enough samples and replicates for any of the subsets.

However, the dispersion/size factor estimation takes significantly longer than anticipated so we thought it would make sense to calculate these first and then apply the subsetting procedure to pull apart specific contrasts. Will this impact the statistical power of our experiment? Is there any difference in subsetting the data before or after calculating dispersions/size factors?

Cheers and many thanks in advance for any and all advice,

Ben