Le lundi 01 octobre 2007 à 10:30 - UM2 - Bât 09 - Salle 331Elodie Brunel
Consider an i.i.d. sample of a random pair (X,Y). We provide an adaptive nonparametric strategy to estimate the conditional density of Y given X = x. We prove that our estimator reaches optimal rates of convergence in a context of anisotropic function classes. We prove that our procedure can be adapted when the response Y is a positive censored random variable i.e. when only Z = min(Y, C) can be observed, for a censoring variable C independent of the pair (X,Y). Simulation experiments illustrate the method. This is a joint work with F. Comte and C. Lacour.