Correlational effect size benchmarks.

Area of Science:

• Psychology
• Behavioral Science
• Quantitative Psychology

Background:

• Effect sizes are crucial for scientific research, aiding in hypothesis testing, power analysis, and interpretation of findings.
• Existing benchmarks for interpreting effect sizes (e.g., Cohen's small, medium, large) may not accurately reflect empirical distributions in applied psychology.

Purpose of the Study:

• To establish empirical effect size benchmarks from a large dataset of correlations.
• To compare these benchmarks against conventional interpretations.
• To provide guidance for research design, power analysis, and intervention evaluation.

Main Methods:

• Extracted 147,328 correlations from the Journal of Applied Psychology and Personnel Psychology (1980-2010).
• Developed a hierarchical taxonomy of variables.
• Calculated effect size distributions at omnibus, domain-specific, and fine-grained levels.

Main Results:

• Empirical effect size distributions showed tertile partitions at approximately one-half to one-third of conventionally intuited values.
• Conventional interpretations of small, medium, and large effect sizes do not align with observed data.
• Significant variation in effect sizes exists across different research domains.

Conclusions:

• The study provides empirically derived benchmarks for effect sizes in applied psychology, necessitating a revision of conventional interpretations.
• These benchmarks can inform more accurate research planning, sample size estimation, and statistical power calculations.
• Findings support the evaluation of research domain advancement and the identification of fruitful areas for moderation analysis and Bayesian applications.