Mixed modelling of interurban trips by coach and train

A stated preference (SP) experiment for interurban trips by coach or train was designed and programmed on portable computers with the aid of specialized software. Trip characteristics and socio-economic information about each individual were also gathered. Four variables were considered: travel time, fare, comfort and reliability. Data for 112 individuals were collected, 14 of which always preferred train; this behaviour was attributed to safety and comfort reasons and therefore the latter were treated separately because of their peculiarity. The data for the remaining 98 cases had ten choices available for each individual; thus we had a final sample of 980 observations for the calibration stage. The next phase of the project was to add revealed preference (RP) data which came from two sources: data collected during the SP experiment (i.e. 98 individuals) and another sample taken as part of a previous research project conducted by the Universidad de Chile (407 individuals). The complete data set was used to examine several issues in the mixed estimation of RP/SP models, such as: appropriate model structure, structural variations between the estimation and forecasting stages (including the role of mode specific constants and inertia-like variables), and influence of the relative proportions of RP and SP data in the modelling results. Our findings, particularly in terms of subjective values of time, are compared with previous studies both in Chile and elsewhere.