I don't think he's commenting on statistical analysis, but rather saying that a scientist should follow the clues where ever it may take them, instead of looking for the clues that will get them where they want to be. And of course they fuck up, too.