OPTIMAL KERNEL WEIGHTS UNDER A POWER CRITERION

We develop an approach to choosing optimal kernel weights for nonparam etric estimation of the slope of the regression function which uses ma ximization of power, rather than minimization of integrated mean squar ed error (IMSE), as its optimality criterion. This power criterion lea ds to optimal kernel weights whose derivation is simpler than under ot her criteria and which provides an intuitive understanding of the natu re of the optimality.