Bakker and Wicherts (1) checked the consistency of reported test statistics, degrees of freedom, and and p-values in a random sample of psychology journals. Of 281 published articles 18% included incorrect reporting of statistical results. For about 15% of the articles the statistical conclusion was incorrect, i.e. at least one statistically significant result became statistically insignificant after recalculation, or vice versa.
Errors were more common in low-impact factor journals than in high-impact factor ones, and the errors were often in line with the authors expectation.
Petrocelli et al. (2) quantified articles reporting standard use of single-mediator models in three high-impact journals in personality and social psychology during 2011 to quantify the inaccuracies in this area.
Of the presented 156 models more than 24% failed an equivalence test (i.e., ab = c – c’). This suggests that regression coefficients from mediation analyses frequently are misreported.
Common sources of errors and recommendations for enhanced accuracy in reports of single-mediator models are presented by the authors.
1. Bakker M, Wicherts JM. The (mis)reporting of statistical results in psychology journals. Behav res methods 2011;43:666-678.
2. Petrocelli JV, Clarkson JJ, Whitmire MB, Moon PE. When ab ≠ c – c’: published errors in the reports of single-mediator models. Behav Res Methods 2013;45:595-601.