Is there a two-way multiple range test that I could get SAS to do in some way or another? I would prefer it be Bayesian, since the one-way Bayesian (Waller-Duncan) is less messed up than the frequentist methods if a single sample is much larger than many other samples. (Frequentist methods can have a real problem if one has a system where the control has a mean signal of 1, another sample has a mean signal of 100, and other signals are around 10 through 40. The frequentist methods keep saying that only the highest sample is significant, but if we had simply not happened to have found that particular clone on that particular experiment, it would assign significance to the lower samples. Waller-Duncan is able to distinguish pretty much the same groups with or without the extremely high sample.
This should have been in GQ (although the reaction some people have to Bayesian methods makes me wonder if GD isn’t the right place, sometimes).