Hello!
I am conducting a meta-analysis looking at individual means from each study. Specifically, there are no binary relationships, no treatment or control groups etc.
Essentially, I am pulling a mean score (along with sd, variance etc) from each study using a random effects model. However, I am a bit confused about if I am creating a cohen's d or hedge's g given that there is no " mean difference" but one descriptive mean. I am assuming I can simply add these raw scores (all from the same measure) into the meta-analysis software to test for homogeneity etc but if I chose to add studies using other similar measures how do I standardize them to begin further analysis? Thank you!!
It seems like you're doing a meta-analysis of "single-arm" studies? Often, this approach is used to pool the prevalence of some phenomenon. Because there is only one-arm, variance estimates are often higher. For this reason, some variance-stabilizing procedures, such as the Arcsine transformation, can be applied to help adjust for this. For more information, here is a nice article: PMID: 23963506 (DOI: 10.1136/jech-2013-203104).
While there may be many more in-depth answers on how to handle the statistics, I would encourage you to review your articles to see if you indeed have sufficient data to conduct analyses. One always needs to answer the question of - am I forcing this, or should it be a systematic review. The latter is not a failure, it is a true finding. Good luck. The previous answers may be of more technical assistance. But - I remind you of this because you do not want to submit your publication only to have a reviewer dispute the basis of your analyses.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Meta-Analysis