Nonparametric regression estimation in models with weak error's structure (Q811055)

The authors consider the problem of nonparametric estimation of the regression function and its derivative. The main purpose of the paper is to propose an estimator which has ``good'' asymptotic properties under weak conditions on the error structure, especially the assumption of independence, which is not realistic in many practical situations, is avoided. Under the additional assumption of repeated measurements at the controlled variables the authors consider an estimator of local weighted average type. Assuming that the errors form an \(\alpha\)-mixing sequence and under weak conditions on the weights they prove the uniform mean square convergence and the asymptotic normality (at a fixed point) of these estimates. Furthermore, considering the estimates as a random element on the space of real-valued continuous functions the weak convergence to a Gaussian process is investigated.