Evidence that's built on "statistical correlation" is very weak. A correlation coefficient of 0.6 or less is practically non-predictive. The people in this area need to use better mathematics.