Nobody knows what rskr is debating about. But thanks, sdriscoll for chiming in; I didn't expect anybody to be still reading. ;-)
Just to clarify two points in case somebody reads through this mess:
- Independent filtering is not restricted to microarrays. The paper derives and proves a general result, and merely illustrates it with microarray data. Our application here does not follow that example but uses the general result. (The paper filters on variance, in this thread we use mean.)
- It is in fact true that you can use the p values from the DE test to chose, sort of post hoc, the filter threshold, as long as the filter criterion fulfills the permutation-invariance condition described in the paper. Intuition might make one be sceptical here, so read the proof if needed.
Just to clarify two points in case somebody reads through this mess:
- Independent filtering is not restricted to microarrays. The paper derives and proves a general result, and merely illustrates it with microarray data. Our application here does not follow that example but uses the general result. (The paper filters on variance, in this thread we use mean.)
- It is in fact true that you can use the p values from the DE test to chose, sort of post hoc, the filter threshold, as long as the filter criterion fulfills the permutation-invariance condition described in the paper. Intuition might make one be sceptical here, so read the proof if needed.
Comment