Resumen
In this paper, first we propose sharp sampling schemes for nonparametric density estimation for discretely observed continuous time processes. Next, since such samplings depend on two coefficients: ?0, r0 linked with regularity of the underlying sample paths and the density, respectively, we give and study the pointwise asymptotic behaviour of an adaptive kernel estimator in the case of known ?0 and unknown r0. Finally, we propose a procedure in the case of both coefficients unknown.