Hazard-based duration models have been successfully implemented to study event durations across many disciplines. This paper focuses on integrating – for the first time, to the authors’ knowledge – the hazard-based duration modelling method into a novel bivariate framework while accounting for the cross-equation error correlation, endogeneity, unobserved heterogeneity, and unbalanced panel effects, by employing correlated grouped random parameters. The developed framework provides the flexibility of using appropriate, case-specific distribution of the hazard function for each duration. Greater explanatory power is achieved through estimation of panel specific correlated random parameters, which can account for the interaction between the captured unobserved effects and their impact on durations. For demonstrative purposes, travel time and travel distance for trips in the year 2017 and made by household members from the Miami metropolitan area, FL, are modelled using the proposed method. The results show that using different distributions significantly affects the overall statistical fit, forecasting accuracy, and the interaction of error terms within the models.

基于危害的持续时间模型已成功实施，以研究许多学科的事件持续时间。据作者所知，本文首次将基于危害的持续时间建模方法整合到一个新的双变量框架中，同时考虑了交叉方程误差相关性、内生性、未观察到的异质性和不平衡的面板效应，通过采用相关的分组随机参数。开发的框架提供了灵活性，可以为每个持续时间使用适当的、特定于案例的危险函数分布。通过估计面板特定的相关随机参数，可以实现更大的解释力，这可以解释捕获的未观察到的效果及其对持续时间的影响之间的相互作用。出于演示目的，使用所提出的方法对 2017 年来自佛罗里达州迈阿密都会区的家庭成员进行的旅行的旅行时间和旅行距离进行建模。结果表明，使用不同的分布会显着影响整体统计拟合、预测准确性以及模型内误差项的交互作用。