Likelihood analysis is proposed for a multiple regression model when some explanatory variables are measured with error and the measurement error variance is function of the true variables. The method is applied to data concerning the achievement levels in statistics of students attending the University of Udine and it is compared to the results obtained with the Simex method.
