The multiple regression models for the three clusters and the four clusters were also estimated. The R-squares of both models exceeded 0.6, with the exception of the fringe area (cluster A of the three clusters) and the outer suburbs and periphery area (clusters B and A of the four clusters). The results of both models show that the standardized coefficients of ‘density’ are larger than that of ‘diversity’ in the CBD areas and the fringe and periphery area, and the opposite is true in the subcentral area and inner suburbs.
This outcome supports the research assumption that the ridership of the station areas at both ends of the urban structural hierarchy are affected more by ‘density’ than by ‘diversity’, whereas the ridership of the in-between station areas is influenced more by ‘diversity’ than by ‘density’. This research finding suggests that the policies of the TOD planning factors may affect subway ridership if they are applied properly to the clusters of the city based on the urban structural hierarchy.
Some fine distinctions can be identified by taking a closer look at the estimation results of the models. Although the results of the periphery areas of the four clusters are identical to the above-mentioned one, the fringe areas of the three clusters present a different result: ‘diversity’ is more influential than ‘density’. This finding may be explained by the regional hub footholds as the transition zone (Kim, 2008). The outer suburbs of the four clusters, which belong to the fringe area in the case of the three clusters, have features similar to the inner suburbs. Consequently, clustering the station areas of the SMS into four clusters would be useful to estab- lish a proper land use plan for the station areas to increase their subway ridership.
In general, ‘bus’ influenced all station areas without exception. Therefore, subway ridership would be increased by an additional bus line in any station area.