I am conducting a meta-analysis to explore the differences between two groups on a scale. Studies report several subtests and I want to analyze the mean differences of these subtests.
My question is should I perform several meta-analyses on each subtest separately or combine all effect sizes in a meta-analysis like independent studies (such as study1-a, study1-b) and perform a moderator analysis for subtests using a dummy variable?
Thank you!
Dear Furkan
Based on the information you have provided, If your aim is to analyse the mean difference between two groups on each subtest, then you should carry out a meta-analysis for each of these subtests seperately. I hope this was helpful.
Regards Jesper
It also depends on whether the subtests are based on the same data source, e.g., same sample/respondents. If they do, you need to take the dependencies between those data points into account and this requires a multilevel model. You cannot assume that study1-a and study 1-b are similarly independent as study 1 and study 2 if both subtests of study 1 are based on same source. For example, this could be the case if they represent two outcomes of interest.
A multilevel model can be constructed easily in r using metafor (or meta). use random=~1|Study/outcome or similar. http://www.metafor-project.org/doku.php/analyses#multivariate_multilevel_meta-analysis_models
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Meta-Analysis