So sometimes I forget that the design matrix I have in mind for my analysis doesn't make sense, and I run a differential analysis anyway. If I would do this with
limma for example, I would get an error telling me I have non estimable parameters or something similar. Here I made the mistake running dmrseq, and it ran the analysis without problems, until I got the DMR list with the
qval columns as NAs.
I guess technically generating DMRs is independent of the design, and this is really only used to generate the permutations and assign a p-value.
Do you think it's worth adding a check for this?