Assessing Risk-Taking in a Driving Simulator Study: Modeling Longitudinal Semi-Continuous Driving Data Using a...

Area of Science:

• Traffic Safety Research
• Behavioral Psychology
• Statistical Modeling

Background:

• Signalized intersection management is a key indicator of risky driving behavior in simulator studies.
• Previous research suggests passenger influence significantly impacts teenage driving decisions.
• Longitudinal, semi-continuous driving data with zero-inflation presents analytical challenges.

Purpose of the Study:

• To investigate the effect of passenger risk-attitude (risk-accepting vs. risk-averse) on teenage male drivers' intersection risks in a simulator.
• To introduce and validate a novel statistical model for analyzing complex, two-part longitudinal driving data.

Main Methods:

• A randomized trial design was employed with teenage males exposed to either a risk-accepting or risk-averse passenger.
• A two-part regression with correlated random effects model (CREM) was utilized, combining logistic regression for yellow light stops and linear regression for red light intersection time.
• Statistical simulations were performed to assess the CREM's power and efficiency compared to alternative methods.

Main Results:

• Teen drivers with risk-averse passengers demonstrated a higher proportion of stopping at yellow lights.
• Drivers exposed to risk-averse passengers also spent longer in the intersection during red lights when they did not stop.
• The CREM approach showed superior statistical power in most simulated scenarios.

Conclusions:

• The correlated random effects model (CREM) is an efficient and powerful method for analyzing complex longitudinal driving simulation data.
• Passenger influence, specifically risk-attitude, significantly affects teenage drivers' intersection decision-making.
• Findings support the use of CREM for future driving behavior research and provide guidance for sample size determination.