Uma aplicação do problema da estimação em modelos de regressão com erros nas variáveis

Abstract

The demand for better control of measurement procedures has intensified, leading researchers to seek new techniques to obtain better results. And when it comes to measurements, the help of some measuring instrument is necessary. However, measurement errors are present when these instruments are used. Thus, the objective of this work is to show the problem that exists when the predictor variable of the regression model presents some type of error in obtaining the data, causing the classic regression model to become inaccurate, and consequently, the regression model with error in the variables is appropriate for this case. As an application, data referring to the body mass index and waist circumference of 80 people are used. As the samples are subject to errors, the most indicated is the use of the model with errors in the variables (functional), because in the estimation of its parameters, the errors are taken into account, the presence of measurement errors in the variable (X), directly influence the accuracy of the models' estimators, meaning that as the error linked to the independent variable increases, the standard error also increases.