In the course of my usual complaining about a GWAS-- never mind-- it occurred to me that the core of my problem is the lack of an appropriate null model. /1
Usually when one conducts a study, there is a possible null result that will lead you to say, "Oh well, it didn't work." In a well-functioning scientific world that result is still publishable. What would such a result look like for GWAS? /2
It seems for most investigators, that would be h2=0, no significant SNPs, no genetic correlations with anything. The problem is, THAT RESULT WILL NEVER HAPPEN /3
The first and fourth laws of BG, plus the phenotypic null hypothesis, combine to *guarantee* that there is some h2, some SNPs, some genetic correlations that look like the phenotypic correlations, for *everything*. This should be the null model. /4
If that is all that happens, the response should be, "Oh, well." All talk of "genomic discoveries" should be avoided. But instead researchers have learned that they have a route to a guaranteed result, so they do it over and over. /5
It's exactly like twin studies, where 25 years ago researchers discovered that rMZ is always > rDZ. A guaranteed result, even though it had essentially 0 substance. That is why I wrote the 3 laws paper. Meet the new genomic revolution, same as the old genomic revolution. /end 
