On an inequality of Chernoff (Q1064650)

Let X be a standard normal random variable and let \(G: R\to R\) be a function which is absolutely continuous with respect to Lebesgue measure with Radon-Nikodym derivative g. The inequalities \((Eg(X))^ 2\leq var G(X)\leq E(g^ 2(X))\) are valid. The right inequality is proved by \textit{H. Chernoff} [ibid. 9, 533-535 (1981; Zbl 0457.60014)]. The left inequality is proved by \textit{Th. Cacoullos} [ibid. 10, 799-809 (1982; Zbl 0492.60021)]. The author generalizes these inequalities. In particular, he proved the following Theorem. Let \(\mu\) be a \(\sigma\)-finite measure on (R,\({\mathcal B})\) and X a random variable with density f with respect to \(\mu\). Let \(\chi\) : \(R^ 2\to R\) be a measurable function such that for \(\mu\)-almost all \(x\in R\) the function \(\chi\) (x,\(\cdot): R\to R\) does not change sign. Furthermore, \(g: R\to R\) is a measurable function such that \(G(x)=\int \chi (x,y)g(y)d\mu (y)+c\) is well-defined for some \(c\in R\). Finally, \(h: R\to R\) is a nonnegative measurable function such that \(H: R\to R\) is well-defined by \(H(x)=\int \chi (x,y)k(y)d\mu (y).\) If \(\mu (\{x\in R| g(x)\neq 0,f(x)h(x)=0\})=0\), then the inequality \[ var_ fG(X)\leq E_ f\{g^ 2(x)(f(x)h(x))^{-1}\int \chi (z,X)H(z)f(z)\quad d\mu (z)\} \] holds.