Page 235 - 360.revista de Alta Velocidad - Nº 6
P. 235

Impacts of station accessibility and regional heterogeneity on HSR ridership






                                              Table 3. Results of model estimation



                               Variables                       Model 1                       Model 2



                    Y :Ridership                      Coef.    Std. Err.   pvalue    Coef.    Std. Err.  pvalue
                      it
                                       population     2.714      0.503     0.000     4.280      0.499     0.000

                                     Car ownership  -6.651       0.439     0.000     -4.817     0.389     0.000
                     Socioeconomic
                       variable by    Road length     -0.756     0.149     0.000     -0.359     0.128     0.005
                         region
                                       Fuel price     0.199      0.058     0.001     0.186      0.049     0.000

                                     Business scale   0.235      0.109     0.031     0.295      0.091     0.001

                                          year        0.224      0.008     0.000     0.174      0.007     0.000
                      Time dummy
                        variable        Summer        0.060      0.234     0.012     0.061      0.021     0.004
                                         season
                                         Station
                                        location                   -                 -2.734     1.191     0.022
                      HSR station         Bus
                      accessibility   accessibility                -                 0.296      0.018     0.000
                        variable
                                          Rail                     -                 0.053      0.019     0.005
                                      accessibility

                                Constant             472.113    15.575     0.000    385.896    14.035     0.000
                                         Region       1.610      0.598       -       1.921      0.607      -
                        Random          identity
                         effects      Time identity  0.053       0.101       -       0.051      0.009      -
                      parameters
                                      sd(Residual)    0.182      0.005       -       0.152      0.004      -
                             Log likelihood                    160.829                        289.798

                            Wald chi square                    2771.64                        4307.17

                               Prob>Chi2                         0.000                         0.000




                   In order to re-verify that the elements of accessibility are important and to find out which
                   model is more appropriate to explain HSR ridership, we tried to compare Model 1 and 2 by
                   statistical method. Burnham and Anderson (2004) note that the first step for model selection
                   would be to establish a selection criterion, such as the Akaike information criterion (AIC) or
                   the  Bayesian  information  criterion  (BIC). AIC  is  an  usual  selection  criteria  which  were  well
                   adopted in past studies, however BIC also has been suggested in recent studies. One advantage
                   of the BIC over traditional hypothesis testing is that it has good properties under conditions of
                   weaker regularity compared with the likelihood ratio test (Roeder et al., 1999; Kim et al. 2013).
                   In addition Keribin (1998) demonstrated that under certain conditions, the BIC consistently

                   International Congress on High-speed Rail: Technologies and Long Term Impacts - Ciudad Real (Spain) - 25th anniversary Madrid-Sevilla corridor  233
   230   231   232   233   234   235   236   237   238   239   240