Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE credits
In the thesis, a proposed route choice model is tested on the empirical observation data.
The model gives link specific errors to the links in the network, which allows natural
correlation structure among paths. Indirect inference method is used to consistently
estimate the route choice model, and a logit model is chosen to be the auxiliary model.
The observation data with multiple OD pairs was collected in the Borlänge, a city located
in the middle part of Sweden. Two main experiments are carried out, one with only one
attribute, length to be specific in the route choice model and the other with two attributes,
namely length and speed bump dummy. The final estimates of parameters have positive
signs as anticipated and the magnitudes are reasonable. In addition, judging from the
estimated binding functions, parameters in the cost function have negative effect on
utility in the logti model and the parameter in the logit model are more apparently
affected by its indirectly inferred parameter in the route choice model than the others do.
Some other trials of the model are also carried out, but the estimates are not satisfying.
This may be due to lack of attributes in the route choice model and further suggestion on
better defined model is given out. A Monte Carlo experiment is also carried out to test
the efficiency of model with unfixed scale parameter to the random term in the route