Researchers publishing in the PNAS USA have found that there are significant relationships for two out of six genes they tested amongst friends, after accounting for age, race, sex and common ancestry.
Is this perhaps another example of a great statistical joke being played out at the highest levels? Or perhaps it is a real effect? The fact that we can’t be sure, even after having performed elaborate statistical tests, is probably the best indictment of the flawed methods in statistics, and the flawed ways in which studies are reported and accepted for publication.
Quick question: given the effect size, how many genes would they need to test in order to reliably come up with two spurious results? More than six.
Professor Adam G. Dunn 