A methodology for evaluating the precision of ellipsometrically determined model parameters (such as thickness) as a function of the measured data set (wavelength(s) and angle(s) of incidence) is presented and experimentally validated. This methodology is used to determine acceptable angle of incidence ranges for ellipsometric measurements and to study the merits of including multiple angles in the data set. Simulations and experimental results are presented for rotating analyzer ellipsometer and rotating compensator ellipsometer systems, using examples of thin and thick films on silicon substrates. A rigorous noise model is used to predict the ellipsometric measurement noise and Monte Carlo simulations are used to determine the parameter precision for a given data set. (C) 2003 Elsevier B.V. All rights reserved.