[R-meta] Dear Wolfgang
juhyung2 @end|ng |rom @t@n|ord@edu
Fri Sep 25 19:17:29 CEST 2020
I hope you are well.
I wanted to make a very basic inquiry about mixed-effect models with studies that have multiple treatment groups in a single experiment. I am more familiar with traditional meta-analysis where in each experiment, only a control and a single treatment group (perhaps the highest, or lowest treatment) are compared to generate effect sizes.
However, is it possible to generate multiple effect sizes (based on comparison of a single control with different treatment levels in a single experiment) from such individual experiment and run mixed-effect models to generate grand mean effect size. In this specific analysis, treatments are continuous variable (pH level). I suspect that this, of course, violates independence among effect sizes, but by applying the covariance-variance matrix to account for the shared control group, is it legit way to run the mixed effect analysis this way.
The reason we want to do this is because we want to run a moderator test whether pH level might influence effect sizes, and also because there are not many publications that we can include in this analysis.
I would deeply appreciate your insights, Wolfgang!
[[alternative HTML version deleted]]
More information about the R-sig-meta-analysis