...I don't want to deal with this bullshit.
We determined today that the data I was supposed to combine will likely be influence by technical effects (still have to check because I'm having a bitch of a time getting everything to talk to each other). But... what's the point in combining this shit if I have to go through another 100-1000 CPU hours of processing and they still won't have enough samples to be significant.
On the other hand, I've decided to determine how many samples I need to recapture a larger distribution using bootstrapping. Because I'm tired of having people go "Yeah, I got you 5 samples, can you analyze them, now?" when I can look at 5 samples and say, "yeah, based on your boxplot, you don't have the full distribution". (I might also be at the point where I'd doing that with published data, too).