Recently traffic models have evolved significantly. An important part of this development has been the change from traditional models (individuals are assumed to have identical preferences and possess perfect information) towards models incorporating stochastic elements (generally referred to as Error Component Models). An example of this is the East Denmark Model (EDM) including stochastic elements to describe delays, differences in users' preferences, lack of knowledge of the public transport supply (network, timetable and connections) and general randomness (Nielsen et al, 2000). Within the estimation phase, models are typically solved for simple binary choices in supervised (guided) experiments (Stated Preference interviews). In applications these elements are included by means of various simulation models (train delays) and statistical distributions from which random numbers are drawn. These types of stochastic terms were included at different levels in the EDM; preference and link stochastics and in addition to this correlation between the stochastic terms. The EDM is of a complex character, where the various stochastic terms have different but overlapping areas of impact. During the validation of the model it was decided to carry out a more detailed and systematic test of the EDM stochastic elements, to gain a thorough understanding of the consequence of their inclusion. The present paper covers the analysis of the EDM's stochastic elements, and their impact on the modelling result. Such analyses are - to the authors' knowledge - not undertaken on similar models. Moreover, EDM may be the most complex model with regards to stochastic assignment published in the literature.
Samenvatting