考虑重复观测之间关联性的信息响应建模

摘要

以高架快速路上可变情报板发布高架和地面道路行程时间条件下改道决策的意向数据为建模对象,进行信息响应的离散选择建模,建立二元Probit横截面模型(CM)和考虑重复观测之间关联性的面板模型(PM),分析两种模型的性能.研究表明:驾龄越长,高架使用频率越高;中等年龄段的驾驶员越愿意改道;地面交叉口越少,节省时间越多,越愿意改道;单位用车驾驶员越不愿意改道.PM的拟合度明显高于CM; PM在模型系数上与CM相差不大,但是其统计推断鲁棒性更好;CM会明显高估驾驶员相关的解释变量系数的t检验值,但是会略微低估出行情境相关的解释变量系数的t检验值.%Urban freeway users' diversion response to VMS that display travel time on both urban freeway and local streets was explored. Stated preference data of diversion behavior was used to develop a cross-sectional binary probit model and a panel binary probit model for identifying factors that influence diversion response. Main findings regarding VMS impacts are travel time saving and drivers' years of driving experience serve as positive factors in diverting;number of traffic lights on the local street, using frequency of urban freeway; being a mid-age driver, and being an employer-provided car driver serve as negative factors in diverting. On the modeling aspect,it was shown that the panel model does not provide substantially different model coefficients but more robust statistical inferences for model coefficients as compared to the cross-sectional model,and the cross-sectional model tends to seriously overestimate t-test values for explanatory variables correlated with drivers (e. G. Demographic characteristics) but slightly underestimate t-test values, for explanatory variables correlated with scenarios (e. G. Travel time savings). The findings have implications for better design and operation of advanced traveler information systems and for future effort on survey design,data collection and model estimation.