I am at the end of the data extraction process in a meta-analysis. The extracted data come in different forms, the more frequent being : (1) mean and variance of experimental (with plant species x) and control plots (without plant species x), and (2) correlation coefficients of the response variable with abundance/availability/percent cover of plant x.
With type 1, I would like to use Hedge's d. With type 2, I can use directly Pearson's correlation coefficient.
Can I, by transforming Hedge's d into a correlation coefficient (which I know can be done), put the two type of data in the same model?
Hello,
If you like me to help, please send me an email to [email protected]
I try my best to help out.
cheers
Amir
Dear Emilio
You can do that as effect sizes can be converted to each other. But I think giving two pooled effect sizes based on difference in means and correlation sounds more prudent.
After all difference in means and correlation ARE different and I am not comfortable with pooling them together.
Ramin
Yes, you should be able to analyze these in the same model - as you say, Hedge's d (for a dichotomous IV and continuous DV) and correlation coefficiencts (two continuous variables) can be directly transformed from one to the other. If you are concerned, you could look at this as a potential moderator variable - though if you found differences you'd want to be careful in interpreting them (e.g., if most studies reporting Hedge's g are experimental designs and most studies reporting a correlation coefficient are not, then that could be the difference). Having said that, this is an empirical question - I wouldn't start of with the assumption they can't be combined - test that assumption.
Thanks everyone ! I have been thinking about all yours answers.
To Keegan: what does it mean ? Good question ! My variable is the consumption by herbivore in presence of neighboring plant. Which is what both effect size types can answer. But I agree this is not a perfect fit.
I will thus do as Steven suggested : using effect size type as a moderator, at least if my sample size is sufficient for this. I think it will, but I still have a lot of work ahead of me to be sure.
Hi everyone,
Some of my recent reading helped me (a lot). For those interested, I suggest this excellent book :
Card, N. A. 2012. Applied meta-analysis for social science research. - The Guilford Press, New York. 376 pp
Dear Emilie,
This is possible using the Hedges'g, which standardizes the individual effects sizes, even if presented differently. However, the use of Hedges´g creates a standardised measurement that not represent a particular unit (in my case: kg/m2, when I used).
Best,
PH
- Badges
- Science method
- Similar topics
- Biological Science
- Ecology