Something went wrong!
Hang in there while we get back on track
Meta-Analysis Effect Size Conversions
Converts effect sizes from three studies using different metrics (correlation r, odds ratio, F-statistic) to Cohen's d for meta-analysis. Includes conversion formulas, assumptions, and ready-to-use methods text.
Input
What you provide to the skill
Need to convert for meta-analysis: Study A r=0.35, Study B OR=2.4, Study C F(1,120)=8.2. Convert all to Cohen’s d
Output
What the skill generates for you
============================================================
META-ANALYSIS EFFECT SIZE CONVERSIONS TO COHEN’S d
| Study | Original Metric | Value | η² (if applicable) | Cohen’s d |
|---|---|---|---|---|
| A | r | 0.35 | — | 0.747 |
| B | OR | 2.40 | — | 0.483 |
| C | F(1,120) | 8.20 | 0.064 | 0.523 |
POOLED: Mean d = 0.584
Range: [0.483, 0.747]
CONVERSION FORMULAS USED:
• r → d: d = 2r / √(1-r²)
• OR → d: d = ln(OR) × √3 / π
• F → η²: η² = (df_b × F) / (df_b × F + df_w)
• η² → d: d = 2√η² / √(1-η²)
ASSUMPTIONS:
• r to d: Assumes bivariate normal distribution
• OR to d: Less accurate at extreme base rates (<10% or >90%)
• η² to d: Valid only for two-group comparisons (df_between = 1)
METHODS TEXT FOR YOUR META-ANALYSIS:
Effect sizes were standardized to Cohen’s d using established
conversion formulas (Borenstein et al., 2009; Chinn, 2000; Cohen,
1988). Pearson correlations were converted via d = 2r/√(1-r²),
odds ratios via d = ln(OR)×√3/π, and F-statistics via η² then
d = 2√η²/√(1-η²). All conversions assume bivariate normality and,
for F-statistics, two-group comparisons.
About This Skill
Calculate and interpret effect sizes (Cohen's d, eta-squared, odds ratios, correlations) with context-specific guidance. Distinguish statistical significance from practical importance and convert metrics for meta-analysis.
View Skill DetailsMore Examples
A/B Test: Large Sample Statistical Trap
Demonstrates the classic 'large sample trap' where statistical significance (p=0.008) masks a very small effect size (Cohen's h=0.018). Provides business impact scenarios at different traffic levels and a decision framework for implementation.
Education Study: Flipped Classroom Intervention
Analyzes a teaching methods comparison where the p-value narrowly misses significance (p=0.052) but the effect size (d=0.40) is educationally meaningful. Demonstrates interpreting non-significant results with substantial effect sizes and comparing to domain benchmarks.