Let phi be a smooth function of k + 2 variables. We shall investigate
in this paper the rates of convergence of estimators of T(f) = integra
l phi(f(x), f'(x),..., f((k))(x), x) dx when f belongs to some class o
f densities of smoothness s. We prove that, when s greater than or equ
al to 2k + 1/4, one can define an estimator (T) over cap(n) of T(f), b
ased on n i.i.d. observations of density f on the real line, which con
verges at the semiparametric rate 1/root n. On the other hand, when s
< 2k + 1/4, T(f) cannot be estimated at a rate faster than n(-gamma) w
ith gamma = 4(s - k)/[4s + 1]. We shall also provide some extensions t
o the multidimensional case. Those results extend previous works of Le
vit, of Bickel and Ritov and of Donoho and Nussbaum on estimation of q
uadratic functionals.