ABSTRACT. We investigate nonparametric estimators of zeros of a regression function and its derivatives and we derive the distribution of design points minimizing the expected width of a confidence interval and the expected variance of the proposed estimator.The main goal of this contribution is to derive the optimal distribution of design points for a nonparametric regression estimator of the so-called zero of a regression function, i.e., the location in which the unknown regression function intersects the horizontal axis.A survey of literature concerning optimal design for nonparametric regression models may be found in [20]. The immediately following discussion [7] suggests that there seem to exist two different general approaches. One approach uses the classical optimal design theory in order to find a finite set of support points with associated weights minimizing, e.g., the sum of variances of the nonparametric regression estimator in some pre-defined points [5] , we aim to find the design minimizing the variability of the empirical zero, i.e., the location at which the nonparametric regression estimator intersects the horizontal axis, instead of minimizing the variability of the nonparametric regression estimator itself. Similarly as in [10], the optimal design is obtained by applying standard calculus of variations. The main advantage of our approach is that prior