LOGLINEAR MODELS AND GOODNESS-OF-FIT STATISTICS FOR TRAIN WAYBILL DATA
HyungEun Lee, Kert Viele · Journal of transportation and statistics · 2001
Counts of carloads of train shipments are effectively described with loglinear models. This paper presents models of counts by origin, destination, and commodity type. Such models can highlight structures in the data and give useful predictions. In particular, there are definite interactions between origin and destination and between origin and commodity, and these models can capture these relationships. Model selection depends on the choice of goodness-of-fit statistic; this paper addresses several issues relating to this choice.