I have been using the package sva to look for surrogate variables in a microarray expression table. I have followed the sva manuels instructions carefully. It looks as though I have four significant surrogate variables and my sva output includes a pprob.b matrix composed entirely of 0s. According to the manual this matrix shows the posterior probability that each gene is associated with the variables of interest. Meanwhile, pprob.gam (which shows the posterior probability a gene is associated with one or more latent variables) are all valuesclose to or equal to 1.
The variable of interest I selected was that of disease status so here's my question:
Does the pprob.b table full of 0s just mean that the sva analysis didn't turn up enough evidence to tie one of the genes to disease status (thus, 0 posterior probabilities), and this does not mean none of the genes are linked to it, just that this particular analysis didnt produce results suggesting that?
Is something off in my analysis and these tables mean I will mean I shouldn't apply this output in adjusted DE analysis.?
In other words, is this a big red flag or a perfectly normal occurrence?