On sensitivity estimation for nonlinear mathematical models

A theorem about decomposition of an integrable function into summands of different dimensions is proved. A Monte Carlo algorithm is proposed for estimating the sensitivity of a function with respect to arbitrary groups of variables.