Lies or Misuse?: Comment on “Lies, Damned Lies, and Statistics (in Geology)”



[1] To demonstrate a concern in geological interpretation after statistical hypothesis testing, writing that “geological hypotheses are never ‘true’—they will always be rejected if lots of data are available,” P. Vermeesch (Eos, 90(47), 443, doi:10.1029/2009EO470004, 2009) considers a null hypothesis H0 of earthquake occurrences not depending on the day of the week. He found that his testing result rejects H0, and he argues that the hypothesis testing does not reveal any geological significance. We argue that his conclusion basically demonstrates a Type I statistical error, where the null hypothesis is rejected despite being true. Because the use of hypothesis testing crucially relies on three criteria—the correct null hypothesis, a plausible probability distribution, and an appropriate testing statistic—one will easily obtain an incorrect interpretation of statistical significance if one of these criteria is not met. Vermeesch's argument does not exhaustively address whether the last two criteria are met and is insufficient to claim that statistically the hypothesis should be rejected.