Estimating and validating the cumulative distribution of a function of random variables: Toward the development of distribution arithmetic (Q1810786)

The authors aim to find the cumulative distribution function (c.d.f.) of \(Y=f(X)\), where \(X\) is a random vector with a known distribution and \(f\) is a continuous function which is monotone in each variable. The method generates a sequence of bounds that converge to the required cumulative distribution function. The idea is to partition the domain space into smaller boxes, then to construct upper/lower bounds and an intermediate estimate of the c.d.f. for \(Y\) given that \(X\) is in that box, and to combine the conditional c.d.f.s into the final estimate. Numerical experiments indicate that this approximation is close to the actual distribution after a few iterations. An example of how the methods works for a non-monotone \(f\) is also given. Application of this constructions include risk analysis and optimization under uncertainty.