I have 15 articles. Each articles measured working memory several times. I want to regress this and find if, for example, mean age, intervention characteristics, etc. played a significant factor in the outcome?
From my research, I would need to conduct perhaps a multivariate meta-regression? Or are there alternate ways of doing this, I have RevMan and CMA tools available.
Hi, Dear Azeem Rathore
Regarding your question (Is it possible to conduct a meta-regression with multiple effect sizes reported?) Please check the following links, I hope useful and clarify your doubt.
Regards,
https://www.meta-analysis.com/downloads/Meta-analysis%20Meta%20regression.pdf
https://stats.stackexchange.com/questions/166964/multilevel-multivariate-meta-regression
I have recently conducted a meta-analysis on working memory. In that study, we focused on working-memory training studies, where there is pre-post data from more than one dependent measure. The pre-post design is not difficult to account for, but accounting for the within sample dependency across several working-memory tests is a bit more complicated. Depending on the amount of data you have, you could probably model this as a multi-level design.
I could even send you my script for the metafor package in R if you would find that useful.
If there are multiple effect sizes being estimated from the same population (i.e., working memory being measured several times), then the effect sizes are non-independent and you have to control for this non-independence. The easiest way to handle multiple effect sizes within a study is to aggregate them.
If you can get your hands on Borenstein, M., Hedges, L. V., Higgins, J., & Rothstein, H. R. (2009) Introduction to Meta-Analysis, they explain the process in Chapter 24. I'm pretty sure this can be done in CMA, and also in R (for a primer on how to aggregate in R: http://www.tqmp.org/RegularArticles/vol11-1/p037/p037.pdf).
I have always suspected that there were ways to to do multi-level designs in meta-analysis like @Jan Antfolk suggested, so I would be curious to hear more about that.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Meta-Analysis