1: \begin{abstract}

2: We consider the nonparametric estimation

3:  of the multivariate probability density function and its partial derivative with a support on

4:  $[0,\infty)$ by dependent data. To this end we use the class of kernel estimators with asymmetric

5: gamma kernel functions. The gamma kernels are nonnegative. They change their shape depending on the position

6: on the semi-axis and possess good boundary properties for a wide class of densities.

7: The theoretical asymptotic properties  of the multivariate density and its partial derivative estimates like biases, variances and covariances are derived.

8: We obtain the optimal bandwidth selection for both estimates as a minimum of the mean integrated squared error (MISE)

9: assuming dependent data with a strong mixing. Optimal rates of convergence of the MISE are found.

10: \begin{keywords}density derivative; multivariate dependent data; gamma kernel; nonparametric estimation

11: \end{keywords}

12: 

13: \begin{classcode}62G07; 62G20; 62H12\end{classcode}

14: \end{abstract}