95% CI around cumulative incidence estimate

I noticed that confidence intervals around cumulative incidence estimates are using 1.96 +/- square root of the variance (I.e. the standard deviation) in the cuminc function of the cmprsk package. Why would we not use 1.96 +/- the standard error of the mean sqrt(var/n). This is some lack of basic understanding on my part and I was hoping someone would kindly explain why use one and not the other.


1 Like

There is no simple mean here, and no simple n, as the effective denominator depends on how late subjects were censored.