ABSTRACT
This expression is obtained by means of a standard b inomia l formula.
If p is small , as it typical ly is i n risk theory applications, this relative
standard deviat ion turns out to be rather large. F o r example, if p is
0.01 and s is, say, 1000 and hence E(c) = 10, the 95% confidence l imits
for the outcome c are 4 and 16 (using the no rma l approximation) .
If a small relative error is required, the sample size has to be rather
