Abstract
Nonparametric recursive kernel estimators of a multivariate data generating process (DGP) are presented and their asymptotic biases, variances and distributions are examined. Weak and strong consistency of these estimators are also proved. Remarks on the choice of the kernel function and the bandwidth function are made. Recursive estimates of the marginal and the conditional DGP are deduced from the estimates of the multivariate density. Finally, an application of these estimates to estimation and specification of econometric models is pointed out.

This publication has 16 references indexed in Scilit: