Comparação de métodos de estimação em um modelo linear simples com erro nas variáveis

Abstract

Although the researcher is very careful errors experiment may be present, therefore, is appropriate to evaluate the reliability of data, mainly when different estimation methods are used. Considering this problem, the objective of this work is compare the performance of two methods of estimation known as “Plug-in” and “Attenuation bias”. These methods are used in the regression model with errors-in-variables techniques. The Monte-Carlo simulations are used for generated a linear regression model errors-in-variables, different sample sizes, residuals distributions and lack of fit were considered. It was possible to conclude that, “Plug-in “ and “Attenuation bias” methods were accurate with R2 =90%, however, these methods have better accuracy for symmetrical distributions.