EXPERIMENTS WITH OPTIMAL SAMPLING FOR MULTINOMIAL LOGIT MODELS

In this paper a recently published method for optimizing the sample used in estimating discrete-choice models is tested. The work is intended to identify and explore the elements that influence the effectiveness of this methodology in designing sampling procedures for estimating logit models. The investigation includes both anaytical and numerical tests. The results indicate that the sample optimization method can improve the accuracy of the resulting estimates, as compared with random sample. (Author)