scientific article; zbMATH DE number 4020088

From MaRDI portal
Publication:3763309





Suppose three square integrable random variables X,Y,Z are given. The problem considered in the paper is how well the conditional expectation \(E(X| Y+\epsilon Z)\) can be approximated as \(\epsilon\to 0\) by expressions of the form \(m(Y+\epsilon Z)+\epsilon b(Y+\epsilon Z)+o(\epsilon)\), where \(m(y)=E(X| Y=y)\) and b(.) is some measurable function. The main result says that under some additional assumptions such approximation can be found iff Y has a density f such that the function c(y)f(y), where \(c(y)=cov(X,Z| Y=y)\), is absolutely continuous with derivative D(cf) such that D(cf)/f is a continuous function on the support of the distribution of Y.











This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3763309)