4.2. Logistic regression models for captive stations
According to Table 5, the correlation between distance and travel time, mode and travel time, distance and InboundOut and further- Away and travelFeeD are 0.5, _0.5, 0.38 0.4, respectively. Therefore travel time, travelFeeD and InboundOut were not considered for model selection. As defined in the introduction, a captive station is the one located at the end or near the end of train line, which means this type of the station has a bigger catchment area and transit users have less choice of other stations. Therefore, fewer variables would influence the station choice. The derived model for captive stations provided evidence for this definition by identifying only two significant variables in the best-fitting logistic regression model (see Table 6). There are 274 records for the captive stations (Table 2), but the sample size for this regression model is 245 with 29 missing records being removed for the purpose of the analysis. Similar to the model discussed in Section 4.1, distance has a negative influence on the nearest station choice. In addition, the further-way station choice suggested by the model has negative influences on the nearest station choice. The captive station attracted transit users who are willing to either drive or take buses to reach a station which is further away from their destination, and the nearest station is chosen for reasons such as seat availability.