Weak consistent moving block bootstrap estimator of sampling distribution of CLS estimators in a class of bilinear models

Abstract

Grahn, (1995) introduced the Conditional Least Squares estimators for the class (I) of bilinear models. Such estimators have a variance which is difficulty to derive analytically. The aim of the present paper is to consider the conditions under which to apply the Moving Block Bootstrap to estimate the variance and we show the weak consistency of the above bootstrap method relatively to the sampling distribution of CLS estimators. An important consequence is to select the length of the Blocks by a theoretical Influence function.