Presumably the limits were computed based on n samples: something like xbar +/- k(n,1-alpha,p)*s, where k-is a 2-sided tolerance factor (or 2-sided equal tailed tolerance factor) based on n samples, 1-alpha confidence, and p is minimum proportion of population captured by the interval. Then no more than 1-p is outside the interval with 1-alpha confidence.
... View more
