为探究铁路旅客对离站方式的行程费用、在车时间、候车时间和车厢拥挤度等服务水平因素的偏好差异及其原因，运用混合Logit模型和潜在类别条件Logit模型从个体和群体层面展开分析。首先，基于铁路旅客出行特点和离站交通方式服务水平，以同行人数、行李负载、行程费用、在车时间、候车时间和车厢拥挤度作为情景变量，通过正交设计法生成意愿调查问卷；然后，在线上和线下开展调查；最后，采用极大似然估计标定模型参数。结果表明：行程费用、在车时间、候车时间和车厢拥挤度是影响离站方式选择的关键因素。混合Logit模型标定结果显示，旅客对行程费用和在车时间具有不同于样本均值的个体偏好差异。潜在类别条件Logit模型将旅客划分为3类：类别1的旅客对在车时间、候车时间和车厢拥挤度最为敏感，而类别3的旅客对这3项服务水平因素的敏感性最低。两模型从不同角度证实了旅客对离站交通方式服务水平的偏好差异及其对离站方式选择的影响，为铁路客运枢纽接驳交通方式配置和服务水平优化提供了参考。

To explore the differences in passenger preferences and the cause of railway passengers′ preferences for the service level factors of departure mode, such as travel cost, in transit time, waiting time, and cabin congestion, a mixed Logit model and a latent class conditional Logit model were used to carry out the research at the individual and group levels. Firstly, based on the travel characteristics of railway passengers and the service level of departure modes, a willingness survey questionnaire was created using orthogonal design method with the number of passengers traveling, luggage load, travel cost, in transit time, waiting time, and cabin congestion as scenario variables. Then online and offline surveys were conducted. Finally, the model parameters were calibrated using maximum likelihood estimation. The results indicate that travel cost, in transit time, waiting time, and cabin congestion are key factors affecting the choice of departure mode. The calibration results of the mixed Logit model show that passengers have individual preference differences for travel costs and in transit time that are different from the sample mean. The latent class conditional Logit model divides passengers into three categories. Passengers of category 1 are most sensitive to in transit time, waiting time, and cabin congestion, while passengers of category 3 have the lowest sensitivity to these three service characteristics. The two models have confirmed from different perspectives the differences in passenger preferences for departure modes of transportation and the impact on departure mode selection, providing references for the configuration and service level optimization of connecting transportation modes in railway passenger hubs.