Widening the Applicability of Confidence Interval Methods: The Propagating Imprecision Approach

Suppose we want to calculate a confidence interval for some quantity F = f(p1, p2,…, pk) derived from k parameters (often but not necessarily proportions), each of which is estimated independently. In the simplest cases, k = 2, and f is some comparative measure such as the difference or ratio of two proportions or the odds ratio. In general, k may be greater than 2, and the relationship of f to its parameters may be non-linear, possibly even non-monotonic.