Why the discrepancy? Is it really possible that many of the studies are false positives? Now this hypothesis can be tested, by determining whether the rate of positive findings is consistent with the reported effect sizes [17]: in other words, we can ask, given what the literature tells us about the effect size, how many positive reports should we expect to find? Ioannidis applied this test to structural brain imaging findings and observed 142 statistically significant findings among 461 studies, while the average power of these studies indicated that we should expect only 78.5 significant findings. This difference was itself significantly different [18•]. A recent meta-analysis investigating the effect of the serotonin transporter on amygdala activation provides the necessary data to test for the excess of false positives in imaging genetics studies [19]. Applying this, we find 11 statistically significant findings when 5.5 are expected (P = 0.027).
