ON THE LIKEHOOD RATIO
AND THE KULLBACK-LEIBLER DISTANCE

GONZALO PEREZ IRIBARREN Centro de Matematicas. Facultad de Ciencias. Universidad de la Republica. Montevideo, Uruguay.

ABSTRACT

In this note the LR is seen in the perspective of the generalized Kullback-Leibler distance. The generalized K-L distance permits to obtain a unified vision of the Maximum Likehood Estimate, mainly in the case of Models, i.e when the underlying distribution in the sampling does not correspond to any parameter value. This is specially interesting when dealing with dependent observations, and show some robustness of the ML method. Likewise the generalized distance leads to extend to the dependent case several well-known properties of the likehood ratio in the independent case.
If you wish a full copy of this article please write to:
nmoller@cmat.edu.uy
Back to PMU6, Table of contents