Converting Correlation to Effect Size

#1
Has anyone had experience of converting various indicators of effect size (students t, odds ratio, correlation etc.) to Cohen's d effect size?

Using the formula described in Borenstein et al's Introduction to Meta-Analysis (2009, chapter 7) seems to give high Cohen's d values [d=2r/(sqrt 1-r^2]. Converting from other indicators of effect size gives more expected values.

Any thoughts or comments appreciated.

John
 

hlsmith

Omega Contributor
#2
I don't have a tremendous amout off experience in this area, but I did try the Borenstein presented conversions once. I noticed they seemed a little off. Perhaps it was with correlations, I don't recall. No more insight than above. Let us know what you find out.
 
#3
Thanks for your comments. I have played with some test data and compared Cohen's d calculated from means and SD to the different methods of converting other indicators of effect size. Converting from the correlation coefficient seems the most variable compared to odds ratio, students t, and means/SD. I will post any further findings.

John