[dtech]

Good points. It's even more than 1 in 20 when you realize that studies does no get published with p > 0.05, so in the "published studies" sample this effect should be even larger.