Dear community,
I am wondering how to synthesize different types of experimental findings in a meta-analysis. In particular, if you are interested in the effect-strength of a specific treatment across multiple studies how do you code studies which have a split sample or multiple interaction effects? For example, one study shows that the treatment is significant for a group of experts, but not significant for a group of laypeople. Intuitively, I would have included both effect sizes separately, because I need to aggregate the effect of the treatment across all experiments. Is there any literature on this problem?
Many thanks!
Ekaterina
I have been part of a study where we ran into that problem, specifically with different subgroups used (e.g. IVF naïve, vs non naive). What we did was to include them separately (as a and b), weighted by sample size of course. The trick is to make sure there is no overlap between the groups.
Hi, maybe you need to specify more your population of interest OR consider both effects separetely so that you can you can perform sensitivity or subpopulation analysis.
Hi I absolutely agree with Janice and Souleymane your best best is to do a sub group analysis.
A good literature or guide with details of this is the Cochrane hand book
https://handbook-5-1.cochrane.org/chapter_9/9_6_2_what_are_subgroup_analyses.htm
https://training.cochrane.org/handbook/current/chapter-10
Hope this helps . Goodluck
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Meta-Analysis