Extension of Stein's lemma derived by using an integration by differentiation technique

Abstract

We extend Stein's lemma for averages that explicitly contain the Gaussian random variable at a power. We present two proofs for this extension of Stein's lemma, with the first being a rigorous proof by mathematical induction. The alternative, second proof is a constructive formal derivation in which we express the average not as an integral, but as the action of a pseudodifferential operator defined via the Gaussian moment-generating function. In extended Stein's lemma, the absolute values of the coefficients of the probabilist's Hermite polynomials appear, revealing yet another link between Hermite polynomials and normal distribution.