Modelos de estrutura de covariância para dados longitudinais considerando o plano amostral complexo

Abstract

In this monograph statistical methods are studied with the support of computational tools with the aim of analysing panel data from the British Household Panel Survey (BHPS). Statistical inference procedures for parameters of mixed models are evaluated considering different longitudinal covariance structures. Comparisons between estimation methods that consider the complex sampling design and those which ignore it are carried out through a simulation study and the behavior of estimators which are based upon adjustment functions is assessed. The following estimation methods are considered: maximum likelihood, maximum pseudo-likelihood and generalized least squares. The performance of the estimators is also evaluated for scenarios which consider data being generated from different probability distributions, including both symmetrical and asymmetrical distributions.