COVARIANCE MODEL WHEN SOME COVARIATES ARE MEASURED WITH ERROR

Abstract

Frequently, in scientific research, the values of the measured variables can be subjected to errors of observation of appreciable magnitude. Ignoring these errors and applying statistical methods designed for variables measured without error, can yield estimators with undesirable properties. Thus, the importance of obtaining statistical methods to analyze the effect of errors in the explanatory variables. Here the methodology of errors in the variables, derived through the maximum likelihood technique, was applied when some of the variables are subjected to error. To achieve this objective, partition of the model and generalized inverses were used. The software system Mathematica was used to solve problems of arithmetical calculations, involved in the derived methodology