For my master's thesis, I am using meta-analytic techniques to explore the relationship between LGBTQ community connectedness and indicators of physical health. I am conducting multilevel meta-analyses in order to include multiple effect sizes from the same articles while accounting for the nested structure of the data. I am using the metafor package in R to run my analyses and I would like to run a separate meta-analysis for each of the four health outcomes that I am investigating. I am new to R and I am wondering if there is a specific command that I can use to calculate four separate overall effect sizes (one for each of the four health outcome variables) and test moderators for each of the four effect sizes separately. So far, I have only been able to calculate one overall effect size for the whole data set.
Hi Kendall,
I would code the data such that you have all correlations in one column but a separate categorical variable which labels the respective outcome.
In R, you draw subsets of the data when running the escalc() function based on a specific entry of the variable.
That allows you to estimate an overall average correlation (actually, I would not call that "effect size" :)) across all outcomes as well as specific correlations.
HTH,
Holger
Hi Kendall
Another package you may want to consider is metaviz, which allows you to conduct subgroup analysis. Unlike metafor, it defaults to fixed effects, but you can specify random effects (only flagging this up because I found differences in results between the two packages until I specified REML in metaviz). What's great about metaviz is it automatically produces a forest plot image, and when you subgroup you can specify different colours for each (very pretty with a good colour scheme). Below is a link to a how-to guide.
https://cran.r-project.org/web/packages/metaviz/vignettes/metaviz.html
Agree with the above. I'd suggest to draw the subsets in producing the actual meta-analytical models, so you still work with the same data set. To make it more specific:
ma.ONE
- Badges
- Science topic
- Similar topics
- Biological Science
- Kinesiology
- Running