Meta-analysis standard errors
I've run an internal meta-analysis on t-tests from 10 different studies I conducted. I had to calculate in R the standard error of the effect sizes (d), which worked well. Now, I'd like to run a meta-analysis on the indirect effect of a simple mediation analysis across the 10 studies, what effect size and standard errors should I use? Any information would be greatly appreciated!
Comments
I'll forward this to our expert!
E.J.
I'm assuming you mean you want to meta-estimate the product c' = a*b in the mediation X --a--> M --b--> Y. The problem of the SE of c' = a*b is that the analytical SE's are unknown, and that's why people usually use a bootstrap to find that (for which you need the raw data). Fortunately most papers will report the bootstrap SE in one form or another (i.e., the SE itself, the confidence interval, a z-value or a p-value); so you could try to extract them from those papers.
Hi Raoul,
Thank you for your response! Yes, I'm trying to calculate c' - the indirect effect. Fortunately, those are all my own studies, so I have all the raw data needed.
So in this case I can just take the " Std. Error" (0.060) and the "z -value" (-0.358) as an effect size to compile the meta-analysis, or do I need to convert them to some other values?
Here's the output of the indirect effect
Estimate Std. Error z-value p Lower Upper 95% Confidence Interval
X → M → −0.021 0.060 −0.358 0.721 −0.139 0.096
Note. Delta method standard errors, normal theory confidence intervals, ML estimator.