Dear All,
I would like to do a meta analysis of prevalences in STATA and was
wondering if I could get some advice.
In a few published meta analyses of proportions, I noted that people
have taken the logits of the prevalences and then back-transformed the
estimates. Is this the best way to proceed? I have a few studies for
which the prevalence is equal to 100% thus the logit transformation
doesn't work. Would I then be best using the empirical logistic
transformation? My only concern there is how I go about backtransforming
the SEs to get the confidence intervals.
Are there other alternatives?
Many thanks for your help.
Kind regards,
Stephanie MacNeill
|