min.sample argument in preDA function about datest

[Russel88]

Hi Morgan,

I do not know of a "correct" way to choose the cutoff, but there are some things to consider when choosing these parameters. First, a feature should have a count > 0 at least in some samples. If it's only found in very few samples you can't really do any statistics anyway. Second, the counts should not be too low, because noise will have a large effect on fold changes with counts close to zero. The challenge then is to choose these parameters such that we remove the features where we don't have enough information, but keep the rare ones that are of potential interest. This choice also depends on your experimental design and number of samples.

Sorry about the late answer. Hope you still find it useful.

Cheers,
Jakob