Power and True Report Probability in the Literature on Mice Nicotine Conditioned Place Preference

A lack of prospective power and use of effect sizes in the literature of various fields have been
revealed and characterized over the years, giving rise to serious doubts on the reproducibility
of many scientific results (Button & al. 2013, Nat.Rev.Neurosci. 14:365-376; Cohen 1962,
Abnorm.Psychol. 65:145-153). To our knowledge, no study has address this problem in the
field of experimental psychopharmacology using animal models. The articles were identified
in PubMed. The sample size, the type of statistical test, its result, degrees of freedom and pvalue were recorded. We then computed the individual and the median prospective powers
for 6 possible effect sizes (Cohen’s d: 0.01, 0.2, 0.5, 0.8, 1.2, 2). The TRP was computed from
the median power, type-I error rate and the plausibility (prior probability). Amongst 139
articles, only 47 met our inclusion criteria for 109 statistical tests. In this sample, 77.57% of
tests were significant. The median powers for small (0.2), medium (0.5) and large (0.8) effect
sizes in F test were 9.56% [IQR 7.96%-11.5%], 34.45% [IQR 24.61%-47.01%] and 70.46% [IQR
52.92%-85.91%]. None of these numbers reached the recommended minimal prospective
power of 80%. A 50% hypothetical plausibility yielded TRPs of 48.9%, 77.5%, and 87.6% for
small, medium and large effect sizes. For a plausibility of 10%, the TRP were 16.1%, 40.8% and
58.5%. These results generalize to a subfield of animal-model experimental
psychopharmacoloy (nicotine CPP in mice) the lack of power reported in the litterature of
several neurobehavioural and psychological disciplines.