This work studies the combination of random effects and measurement errors in the regression models (functional and structural versions). In the proposed model we use the of class elliptical distributions that form a class of generalized families of distributions that preserve the same structure of symmetric as the normal distribution allowing, for example, to accommodate measurement errors through aberrant distributions with tails heavier than normal. We use classic tools ( correctedscore approaches and the EM algorithm) for consistent estimators and their limit distributions. We study the following models: ( 1) functional linear model wi th mixed effects wi th elliptical errors invariables, (2) symmetrical structural model with random intercept and (3) heteroscedastic regression model with errors of measurement for k populations. We also present sorne diagnostic studies using the methods of local influence for assessing the robustness of the estimates for model parameters under different perturbation schemes. We discussed sorne simulation studies and illustrate theresults with real data.Keywords: Measurement errors, distributions elliptical, corrected score, EM algorithm, linear model functional, symmetrical structural model, regression model heteroscedastic, local influence.PFCHA-BecasDoctor en Ciencias106p.PFCHA-BecasTERMINADA