I am running a meta analysis on prevalence data. Participants can belong to one of four groups. Each study produces the percentage of participants that belong to each group (e.g., Group A: 50%; Group B: 20%; Group C: 10%; Group D: 20%).
I am running a meta-analysis with random effects to produce the overall prevalence of membership in each group, accounting for the different sample sizes of each study.
When I do this for each group individually, the prevalence values do not add up to 100. Is this to be expected?