I hope someone more serious is looking into it (or has already done so) but my alarm is checked by the crack-pot flags this guy's raised for me. Most obviously that he tries to demonstrate P(A|B) by testing P(B|A) - if you don't understand basic statistics, your conclusions are suspect.