1: \begin{abstract}

2: We consider a nonparametric regression model $Y=r(X)+\ve$

3: with a random covariate $X$ that is independent of the error $\ve$.

4: Then the density of the response $Y$ is a convolution of the densities

5: of $\ve$ and $r(X)$. It can therefore be estimated by a convolution of kernel

6: estimators for these two densities, or more generally by a local

7: von Mises statistic. If the regression function has a nowhere vanishing

8: derivative, then the convolution estimator converges at a parametric rate.

9: We show that the convergence holds uniformly,

10: and that the corresponding process obeys a functional central limit theorem

11: in the space $C_0(\R)$ of continuous functions vanishing at infinity,

12: endowed with the sup-norm. The estimator is not efficient.

13: We construct an additive correction that makes it efficient.

14: \end{abstract}