RUBIT - A MULTIPLICATIVE CONCEPT FOR CHOICE ANALYSIS IN TRANSPORTATION PLANNING

Traditional random utility models like Logit or Probit are always based on a random utility term which is additional to the objective utility term. This produces some of the well known limitations of the more traditional solutions, e.g. the IIA-property and the inflexibility of the multinomial Logit model or the parameter estimation problems of the multinomial Probit model. For the Rubit model the random utility term is defined as a multiplicator of the objective (or deterministic) utility. It is pointed out that this is more reasonable to describe the real decision situation of travellers. Different types of distribution functions for the random utility term are possible like the Gamma distribution or the Rayleigh function. The easiest and more flexible solution is obtained using the log-normal distribution for the random utility term. Also multiple choice models with larger numbers of alternatives can be treated. Parameter estimation is possible using maximum likelihood techniques. The model concept, in general, does not have the IIA-property. As examples two applications of the model are given. They concern a concept for models of passenger access to airports in Germany. The examples demonstrate the usefulness and the operability of the new concept. For the covering abstract, see IRRD 889300.