ABSTRACT
The conditional variance is estimated by approximating E(X~IX1 _k) by a local polynomial, which is achieved by letting 8i, i = l, ... , T be the minimizers of
Let us close Sec. 2.4 by tying the functionals to the general principles outlined in the introduction. The two statistics mentioned there, in the present context, clearly are Mk(x) and fJkx. They both estimate the same qmmtity PkX under a linearity hypothesis of Mk(x) = pk:c (But it should be noted that in the randomization/bootstrap approach we will rather evaluate Lkfunctionals under the null hypothesis in Eq. (2.3)). In the case of L 1 (M£), Mi(x) and h play the same role, whereas in the case of Lr(M 11 ) only the statistic M t (x) remains.
