A New Piecewise Multinomial Logit Model of Crash Severity with Accommodation of an Unknown Inflection Point

Document Type

Conference Paper

Publication Date

2024

DOI

10.13023/2024.RSS14

Publication Title

2024 Road Safety & Simulation Conference

Pages

2 pp.

Conference Name

2024 Road Safety & Simulation Conference, October 28-31, 2024, Lexington, Kentucky

Abstract

[Introduction] Multinomial logit regression (MNL)-based models have been applied extensively to model the relationship between crash severity and its contributing factors (1). Despite the methodological advancements around MNL, traditional MNL assumes a linear relationship between crash contributing factors and utility with a constant sensitivity among contributing factors. However, employing a constant sensitivity scheme among contributing factors may not reflect the complex relationship between crash severity and contributing factors. To the best of our knowledge, there is still a lack of models that provide modeling flexibility to accommodate asymmetric sensitivities of crash contributing factors. Thus, this study proposes a new piecewise MNL model that can accommodate the asymmetric sensitivites by introducing the concept of unknown inflection points.

Rights

© 2024 University of Kentucky.

"Author retains copyright and is free to reuse the content elsewhere."

ORCID

0000-0002-8191-2786 (Xie), 0000-0003-2808-8852 (Yang)

Original Publication Citation

Yang, D., Zhao, Z., Xie, K., Yang, H., Shen, T., & Jeihani, M. (2024). A new piecewise multinomial logit model of crash severity with accommodation of an unknown inflection point [Conference paper]. 2024 Road Safety & Simulation Conference, Lexington, Kentucky. https://doi.org/10.13023/2024.RSS14

Link to Full Text

Share

COinS