This repository summarizes some work from my PhD thesis describing the independence of the NB mean as a filter statistic. The aim is to allow others (and myself!) to examine and update this work without having to modify the code in the thesis repository. There are several components available:
theory: This contains a proof of the independence of the sample mean to the p-values under normality. It extends this idea to the case of correlated tests under normality. (However, I am told that the notation could use some work, so YMMV.) It also contains some comments about the use of the overall NB mean for filtering.simulations: This contains extensive simulations to demonstrate the independence of the NB mean from p-values computed using edgeR. The use of the NB mean is (very weakly!) justified by analogy to the normal case, using the maximum likelihood estimate of the mean in both cases. This is necessary as an equivalent proof for the independence of the NB mean does not seem to be possible.