We want to do a prevalence meta-analysis in Stata but 2-3 studies are having zero event rates. The standard proportion command excludes those studies from the forest plot. However, we want to include them too and want to calculate their effect size too. Can someone please guide which command to use in Stata?
Given you have the population size and the sample size, you can calculate the probability of getting zero events given any hypothesized event rate in that population. You can find the event rate that gives you at least, say, 50% chance to observe 0 events in your sample and use this as an estimate of the event rate in that population.
This has been addressed before in ResearchGate. See https://www.researchgate.net/post/How_to_calculate_OR_odd_ratio_if_one_of_groups_is_0_in_a_case-control_study
You can just eliminate it from the quantitative synthesis (meta-analysis) or you can add 0.5 events to each group that has 0 events.
Hi Jogender Kumar,
I might have similar problem before and was solved following Victoria's paper (VN Nyaga et al. Arch Public Health (2014) , PMID 25810908). Using metaprop with ftt cimethod(score) should be a good idea. Hope this will help you. I also posted this on https://www.statalist.org/forums/forum/general-stata-discussion/general/1533833-meta-analysis-of-proportion-with-the-problem-of-18-articles-were-excluded-due-to-effect-size-equalled-zero.
Regards,
Xinyi Huang.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Meta-Analysis