Effects of noise in performance comparisons of designs for model identification and discrimination

This paper considers the problem of comparison of fractional factorial designs in model identification and model discrimination from a class of models. We compare performance of designs in model identification and discrimination for a large number of simulated data. The frequency or proportion of correct identification of the best model with the mean square error, the average power value and the average p-value for testing the significance of an important parameter are used as criterion functions in our comparisons. We also study the effects of noise in our performance comparison of designs.