The Journal of Personality and Social Psychology is considered the top journal for personality and social psychology. The journal has three distinct sections with different chief-editors. This replicability-report examines the replicability of research published in the JPSP: Interpersonal Relationships & Group Processes section.

Impact factors are only available for all three sections combined. The average all-time citation of an article in JPSP since 1989 is 100 times per article. SCImago rankings of all psychology journals ranked Journal of Personality and Social Psychology #12 with an SJR-Impact-Factor of 4.8 in 2014.

At present, the replicability-report is based on articles published from 1989 to 2015. During this time, JPSP:IRGP published 1301 articles. The replicability-report is based on 981 articles that reported one or more t or F-test in the text of the results section (results reported in Figures or Tables are not included). The test-statistic was converted into z-scores to estimate post-hoc-power. The analysis is based on 9,978 z-scores in the range from 2 (just above the 1.96 criterion value for p < .05 (two-tailed) to 4.

Based on the distribution of z-scores in the range between 2 and 4, the average power for significant results in this range is estimated to be 51%. This estimate suggests that only half of the published significant results in this range are predicted to produce a significant results in an exact replication study with the same sample size and power (results with z > 4 are expected to replicate with nearly 100%).

The same method was used to estimate power for individual years.

The results show a decrease in power over time. In the past 10 years, all estimates are below the long-term average. The replicability score in 2015 is 50. Based on this score JPSP:IRGP currently ranks 17 out of 27 journals in psychology. This ranking is subject to change when more data from articles in 2015 become available and more journals are included in the ranking.

### Like this:

Like Loading...

Dr. R, I am quite impressed with your work, though admittedly I have not given some of the fundamental However, being at best somewhat competent as an “applied statistician”, I have quite likely an ignorant question. Is no your power estimate the population point estimate? If so, can you not provide a confidence interval around this estimate? Relatedly, for each individual study in which you “estimate” the “true” power, could you not aggregate these as in a meta-analysis and provide a “combined” power estimate with CI?

LikeLike