ALGUNAS ALTERNATIVAS PARA ESTIMAR LA MATRIZ DE COVARIANZA DE UN ESTIMADOR EN UN MODELO DE REGRESION LINEAL

Abstract

Commonly in data analysis with Complex samples, the user uses traditional procedures or standard which are implemented in conventional computing statistical packages, where, by default is assumed that the observations were obtained from an independently and identically distributed sample. Hence, the use of a complex desing or the structure of the population is ignored, generating important distortions in the development of analytical inferences. Based on these arguments, some alternatives for estimating the covariance matrix are examined. In addition, some simple corrections are proposed for obtaining consistent estimators of the standard errors of the vector of the linear regression coefficients.