The Importance of Replication in the Study of Random Bullshit

I would never recommend doing more than like three, four, five different comparisons. Once you start having dozens of comparisons, you have to correct for those because you're going to hit on something. The fact that a few of them are significant doesn't mean anything. Even if you did the adjustment comparing for multiple, you know, adjusting for multiple comparisons and it was still significant, you would still be left wondering, was this a fluke or is this a real phenomenon? And that's where the replication comes in.