Rejoinder: Editorial Policies Regarding Statistical Significance Tests: Further Comments

Abstract

In this response to Robinson and Levin’s comments on Thompson (1996), it is argued that describing results as “significant” rather than “statistically significant” is confusing to those persons most susceptible to misinterpreting this telegraphic wording. Contrary to Robinson and Levin’s view, it is noted that the utility of the characterization of results as being due to “nonchance” is limited by the nature of the null hypothesis assumed to be true. It is suggested that effect sizes are important to interpret, even though they too can be misinterpreted; recent empirical studies of publications indicate that effect sizes are still too rarely reported. Finally, the value of “external” replicability analyses is acknowledged, but it is argued that “internal” replicability analyses can also be useful, and certainly are superior to statistical significance tests regarding evaluating result replicability, because statistical significance tests do not evaluate replicability.