Divergence Empirique et Vraisemblance Empirique Généralisée

In this paper, we generalize the results obtained with the Kullbackdistance (corresponding to empirical likelihood) and Cressie-Read metrics(generalized empirical likelihood) to general discrepancies, for someconvex functions satisfying a few regularity properties. In particular, weintroduce a new Bartlett correctable family of empirical discrepancies, theQuasi-Kullback, out of Cressie-Read family, which possess interesting nitesample properties. We conclude this work with some simulations in the multidimensionalcase for different discrepancies.