1: \begin{abstract}

2: In a multiple testing context, we consider a semiparametric mixture model with

3: two components  where one  component is known  and corresponds  to the

4: distribution of $p$-values under the null

5: hypothesis and the other component $f$ is nonparametric and stands for

6: the distribution  under the  alternative hypothesis. Motivated  by the

7: issue of local false discovery rate estimation, we focus here on the

8: estimation   of   the   nonparametric   unknown   component $f$  in   the

9: mixture, relying on  a preliminary estimator of  the unknown proportion

10: $\theta$ of true null hypotheses. We propose and study the asymptotic properties of two 

11: different estimators  for this unknown component.  The first estimator

12: is a randomly weighted kernel estimator.  We establish an upper bound for its pointwise quadratic

13: risk,  exhibiting the classical nonparametric rate of convergence over a  class

14: of Hölder densities. To our knowledge, this is the first result

15: establishing convergence as well as corresponding rate for the

16: estimation of the unknown component in this nonparametric mixture. 

17: The second estimator is a maximum smoothed likelihood estimator. It is

18: computed through an iterative algorithm, for which we establish a

19: descent property.

20: In addition, these estimators are used in a multiple testing

21: procedure in order to estimate the local false discovery rate. Their 

22: respective performances are then compared on synthetic data.

23: \end{abstract}

24: