Strongly consistent density estimation of the regression residual

Consider the regression problem with a response variable YY and with a dd-dimensional feature vector XX. For the regression function m(x)=E{Y|X=x}m(x)=E{Y|X=x}, this paper investigates methods for estimating the density of the residual Y−m(X)Y−m(X) from independent and identically distributed data. For heteroscedastic regression, we prove the strong universal (density-free) L1L1-consistency of a recursive and a nonrecursive kernel density estimate based on a regression estimate.