A Bayesian finite mixture change-point model for assessing the risk of novice teenage drivers

Li, Qing; Guo, Feng; Kim, Inyoung; Klauer, Sheila G.; Simons-Morton, Bruce G. · 2017 · openalex

DOI: 10.1080/02664763.2017.1288202

archive: archived pipeline: cataloged

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper addresses the challenge of identifying the specific point in time when driving risk significantly decreases for novice teenage drivers, a critical parameter for safety education and graduated driver licensing (GDL) regulations. While previous studies often aggregated data into fixed calendar-time intervals, this research argues that cumulative driving time is a more relevant measure of experience. The study aims to detect these change-points while accounting for the substantial heterogeneity in risk profiles among individual drivers, which existing models assuming identical change-points fail to capture. The authors propose a hierarchical Bayesian finite mixture model (BFMM) for recurrent-event change-point detection. The model assumes that crash and near-crash (CNC) events follow a non-homogeneous Poisson process with piecewise-constant intensity functions. Drivers are assumed to belong to latent clusters, where members of the same cluster share identical intensity rates and a single change-point. The number of clusters is unknown and estimated using the Deviance Information Criterion (DIC). The method was applied to data from the Naturalistic Teenage Driving Study (NTDS), which continuously recorded the driving behavior of 42 novice drivers in Virginia for 18 months using in-vehicle instrumentation. The analysis focused on the 38 drivers who experienced at least one CNC event, utilizing cumulative driving time as the exposure metric. Simulation studies were conducted to validate the model’s performance under various configurations. When applied to the NTDS data, the model identified three distinct clusters of drivers with different risk profiles. The estimated change-points for these clusters occurred at 52.30, 108.99, and 150.20 hours of driving after initial licensure. The results indicated that the overall intensity rates and the patterns of risk reduction varied substantially among these groups. For instance, one cluster experienced a rapid drop in risk shortly after licensure, while others maintained higher risk levels for longer durations. The significance of this work lies in its ability to move beyond aggregate averages to identify subgroups of novice drivers with distinct risk trajectories. By quantifying the specific driving hours at which risk changes for different driver types, the findings provide a more nuanced basis for tailoring safety education and parent management programs. Furthermore, the results offer crucial reference points for policymakers designing GDL regulations, suggesting that a one-size-fits-all approach to driving experience requirements may not be optimal for all teenagers.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.8-27b-gittensor on 2026-08-23; verification: pending re-verification.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• novice drivers
• telematics crash prediction
• learner drivers

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).

• Empirical Findings: crash risk outcomes
• Theoretical Contribution: computational model