Mixed logit applications in travel behaviour research
分析了2004至2024年间混合logit模型在出行行为研究中的应用，发现研究者处理七个关键问题的做法差异大，且近年应用增多但问题处理频率下降，2012年后趋于稳定，并提出了五个建模建议。
In the past 20 years, mixed logit models have become mainstream for analysing choice behaviour. This paper analyses common approaches to capture travel behaviour using mixed logit models. Based on a sample of mixed logit applications from 2004 to 2024, the paper investigates to what extent the applications address seven critical issues in mixed logit modelling. The results show a high variation in how researchers address these issues in applied work. The paper also analyses whether there are time trends in the published literature. One trend is that more applications have used mixed logit models in recent years, but these applications address the seven issues less frequently than previously. However, the downward trend disappears after 2012, where, on average, papers from 2012 to 2015, 2016 to 2017, 2022, and 2024 address the issues to a similar extent. Based on this analysis, the paper suggests five modelling aspects relevant to mixed logit applications. In conclusion, the paper highlights the necessity for transport research applying mixed logit models to consider why it applies a specific mixed logit approach, what kind of distributions it applies, how it addresses systematic heterogeneity, how robust the analysis is, and finally, how to validate the models.