Blogger Neuroskeptic has the scoop.  When I worked at Yale, a senior NIH statistician admitted that few people understood the statistics used in modern neuroscience research - including most of the researchers themselves. 

My suspicion is that much of the sloppiness that goes on in these studies is not noticed simply because so few people can decipher the complexity of the statistics used.  And along with complexity we also have "flexibility" as Neuroskeptic points out.