Parameter estimation methods in covariance model with error in covariate

The present paper approaches the covariance analysis model with one factor and measurement error in the covariate. Accuracy and precision of two estimators suggested in the literature were evaluated through data simulation, for estimating parameters of a regression model with measurement error So called Plug-in method estimates the real value based on the observed ones and then uses the common function for estimating the desired parameter The other estimator known as bias smoother, only performs a bias correction on the usual estimator by computing a factor. Behavior of both estimators was studied under different residual distributions, goodness of fit and sample sizes. It is worth noting that, in covariance analysis model, the high the sample size, the better for accuracy and precision. Results suggest that the Plug-in estimator presented the best performance both for accuracy and precision under normality, for the distinct evaluated situations. When the estimators had been evaluated in the model of ANCOVA with the residues distributed for Gamma, the same ones had gotten the worse performance in relation when they were evaluated by the others distributions.