Consider the following example -
If I am studying (Systematic review and meta-analysis) the catch growth among preterm babies, all the included trials studied singleton pregnancies but one of the trials have studied twin pregnancies and they compared catch up the growth of twins versus singletons. Should I include this study in the meta-analysis? (My concern is twins will have a different pattern of growth which may not be comparable to normal fetus) Will this increase the heterogeneity considerably (fixing a random-effects model will solve this issue)?
Thanks in advance.
1. Include the study if it meets the inclusion criteria of your Systematic Review.
2. Yes, theoretically random effect model is recommended for heterogeneity in included studies
I would exclude the twin pregnancies study (and such exclusion should have been stated in your study design). The random effects model is suitable for studies that estimate different expected effect sizes WHEN the reasons that they differ are unknown (hence random). So I would perform random effects meta-analysis only for heterogeneity that can not be explained, which does not apply here.
Thank you Muhammad M Hammami for explaining the critical issue nicely. For example: heterogeneity and random effect model - "perform random effects meta-analysis only for heterogeneity that can not be explained". It is also my takeaway from the discussion. Thanks
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Meta-Analysis