Also, if you look for statistical anomalies in a random group of people, you just have to pre-screen for 30 variables that you don't mention in your report to get one with p<0.05