Carlos S. Alvarado, PhD, Research Fellow, Parapsychology Foundation

Psychologist Patrizio Tressoldi has just published a paper about statistical power in experimental parapsychological research (Prospective Statistical Power: Sample Size Recommendations for the Investigation of the Main Parapsychological Phenomena. Journal of Scientific Exploration, 2016, 30, 10–15).

Patrizio Tressoldi 4

Dr. Patrizio Tressoldi

Here is the abstract:

“The aim of this paper is to offer a practical guideline for researchers investigating parapsychological phenomena to choose appropriate sample sizes to achieve a statistical power equal to or above 0.80. The availability of different meta-analyses related to different parapsychological phenomena allow a sufficient estimation of the expected effect sizes, which usually range from small to very small. With these measures, it is possible to estimate the numerosity of sample sizes necessary to achieve a level of statistical power that can facilitate the replication of different parapsychological phenomena. I discuss ways to deal with the investigation of phenomena with very small effect sizes requiring very large sample sizes.”

The author concluded that “when investigating ESP with different procedures based on free response protocols, it seems not difficult to achieve a satisfactory statistical power with samples ranging from 20 to approximately 100 participants. On the contrary, for the investigation of ESP using classical forced-choice protocols, it seems quite impossible to achieve a satisfactory statistical power given the necessity to recruit approximately more than 10,000 participants.”

He discussed options to lack of statistical power:

“Among the possible solutions to this problem, one is to try to increase the expected ES by recruiting selected participants. As demonstrated by Baptista et al. (2015) in all free-response protocols, . . . selected participants, that is participants who have experience with these kind of tasks and are very committed to succeed, obtain almost a double ES with respect to non-expert participants.”

“Among the other solutions, we suggest disclosing the problem of statistical power and ignoring p values altogether, focusing more on ESs and their estimation by using confi dence intervals in line with the so called “statistical reform” movement endorsed recently by the editor of Psychological Science (Eich 2014), underlying their importance for meta-analyses . . .”

“To summarize, the take-home message of this methodological paper is: Before speculating about the theoretical reasons underlying the unreliability of evidence of most if not all parapsychological phenomena, we must exclude the possibility that it may be due to the neglect of statistical power.”