A general version of Price's theorem
Abstract
Assume that is a centered random vector following a multivariate normal distribution with positive definite covariance matrix . Let be measurable and of moderate growth, say . We show that the map is smooth, and we derive convenient expressions for its partial derivatives, in terms of certain expectations of partial (distributional) derivatives of . As we discuss, this result can be used to derive bounds for the expectation of a nonlinear function of a Gaussian random vector with possibly correlated entries. For the case when has tensor-product structure, the above result is known in the engineering literature as Price's theorem, originally published in 1958. For dimension , it was generalized in 1964 by McMahon to the general case . Our contribution is to unify these results, and to give a mathematically fully rigorous proof. Precisely, we consider a normally distributed random vector of arbitrary dimension , and we allow the nonlinearity to be a general tempered distribution. To this end, we replace the expectation by the dual pairing , where denotes the probability density function of .
Cite
@article{arxiv.1710.03576,
title = {A general version of Price's theorem},
author = {Felix Voigtlaender},
journal= {arXiv preprint arXiv:1710.03576},
year = {2020}
}
Comments
Accepted for publication in "Journal of Theoretical Probability"