Open campanam opened 8 months ago
The issue is that there is no way to average the C.I.s this way. Better to use a binomial confidence interval for an alternative.
Standard deviation (normal approximation) of bootstrapped interval gets negative values. We can simply weight the 95% C.I. critical values using the same weighting as the mean.
Current bootstrap confidence interval assumes a normal distribution of bootstrapped replicates. Instead, a better approximation would be to use the percentile interval.