Hello!
I am creating a meta-analysis based on hazard ratios given for breast cancer incidence in BRCA1/2 carriers. Would this be fixed or random-effects? I am creating a forest plot using RevMan and after reading the examples for each type I am still confused!
Thanks in advance!
Carrie OH
Usually first we run a Fixed Effect model. Then we run Random Effect model if we see heterogeneity between studies based on I-squared value. This link maybe useful https://www.statsdirect.com/help/meta_analysis/heterogeneity.htm
The fixed-effect model is usually inappropriate. It is only appropriate when all studies were done basically the same way. That rarely happens. It is not appropriate to make a model decision based solely on the degree of heterogeneity. That is an outdated procedure. For an easy-to-read reference, see Introduction to Meta-Analysis by M. Borenstein et al (2nd edition).
Agree with Gordon here -- fixed-effect models are typically not adequate (see https://www.meta-analysis.com/downloads/M-a_f_e_v_r_e_sv.pdf)
- Badges
- Science topic
- Similar topics
- Counseling Psychology
- Confusion