This paper introduces a dynamic structure, which incorporates both space and time dimensions within a discrete choice multinomial probit model framework; it gives extensive mathematical-statistical details. In 1997, RA Garrido and HS Mahmassani estimated a space-time multinomial probit model for a freight forecasting problem, where the geographical region analysed was the US state of Texas. Three model specifications were considered: one for summer time, one for spring time, and one for winter time. The dependent variable was the generation of freight transport demand in the short term. Several sociodemographic variables, including population, employment rate, and average wages, were considered in specifying the potential function. These variables were specified as a combined variable, formed by a weighted summation of the sociodemographic attributes. The paper presents some results obtained by applying the multinomial probit model to this problem.
Abstract