Correlational effect size benchmarks.

• DOI: 10.1037/a0038047
• Publication Date: 2014-10-14T17:35:45Z
• Authors: Frank A. Bosco; Herman Aguinis; Kulraj Singh; James G. Field; Charles A. Pierce
• Keywords: Generality; Statistical power

Effect size information is essential for the scientific enterprise and plays an increasingly central role in process. We extracted 147,328 correlations developed a hierarchical taxonomy of variables reported Journal Applied Psychology Personnel from 1980 to 2010 produce empirical effect benchmarks at omnibus level, 20 common research domains, even finer grained level generality. Results indicate that usual interpretation classification sizes as small, medium, large bear almost no resemblance findings field, because distributions exhibit tertile partitions values approximately one-half one-third those intuited by Cohen (1988). Our results offer can be used planning design purposes, such producing better informed non-nil hypotheses estimating statistical power sample accordingly. also useful understanding relative importance found particular study relationship others which domains have advanced more or less, given larger phenomenon. Also, our offers about investigation moderating effects may fruitful provide likely facilitate implementation Bayesian analysis. Finally, practitioners use evaluate effectiveness various types interventions.