ABOUT THE DISCRETIZED MAXIMUM LIKELIHOOD ESTIMATOR
Abstract
The most well-known and used statistical estimation procedure is probably the Maximum Likelihood. Unfortunately, apart from rather elementary situations, the analysis of its performance requires complex probablistic tools, namely Empirical Process Theory. Examples can be found, for instance, in the books by Ibragimov and Has'minskii [5], van der Vaart [9], van de Geer [3] or Massart [8]. Our purpose here is to describe a discretized version of the method, which is not so well-known, but with the advantage that one can derive its non-asymptotic performance from elementary tools. We shall explain the method, give its performance and provide several illustrative examples.
Domains
Statistics [stat]
Origin : Files produced by the author(s)