Valid point, but this concern should be somewhat alleviated by the massiveness of the study. I imagine that in 360,000 people, "pure correlation" should be rela

Valid point, but this concern should be somewhat alleviated by the massiveness of the study. I imagine that in 360,000 people, "pure correlation" should be relatively uniformly-spaced, meaning that there's enough people for all "correlative" behaviors to be statistic noise.