Three hundred thirty-five published articles from six medical journals that contained "negative" results were examined to determine the probability of detecting "small," "medium," and "large" therapeutic improvements. The mean probability of detecting a small effect size was.138; for a medium effect size, the probability was.387; and for a large effect size, it was.614.
(Arch Intern Med 1981;141-1307-1310)
Reed JF, Slaichert W. Statistical Proof in Inconclusive 'Negative' Trials. Arch Intern Med. 1981;141(10):1307-1310. doi:10.1001/archinte.1981.00340100063016