Stima via simulazione e calibrazione: confronto fra stimatori alternativi di equazioni differenziali stocastiche uni-dimensionali

Abstract

A simulation based approach is used to compare the performances of some estimators of the drift parameters of selected one-dimensional Stochastic Differential Equations (SDE), namely a maximum likelihood estimator, a naive estimator and an indirect estimator. These estimators are tested on the most classical models of the recent financial literature and on a classical SDE of population dynamics. First, the need for well generated data is evidenced. Second, our results clearly show the role of an indirect estimation procedure as a general strategy to be used in all those circumstances in which the optimal experimental conditions are not satisfied.