This paper is from Session 6B: Probability and p-values — probing the problems

which comes under Topic 6: Innovation and reform in teaching probability within statistics

The interpretation of effect size in published articles

Presenter

• Rink Hoekstra (University of Groningen, The Netherlands)

Abstract

Significance testing has been criticized, among others, for encouraging researchers to focus on whether or not an effect exists, rather than on the size of an effect. Confidence intervals (CIs), on the other hand, are expected to encourage researchers to focus more on effect size, since CIs combine inference and effect size. Although the importance of focusing on effect size seems undisputed, little is known about how often effect sizes are actually interpreted in published articles. The present paper will present a study on this issue. Interpretations of effect size, if they are presented in the first place, are categorized as unstandardized (content-related) or standardized (not content-related). Moreover, the interpretations of effect size for articles that include a CI will be contrasted with articles in which significance testing is the only used inferential measure. Implications for the current research practice are discussed.