A two-part mixed effects model for cigarette purchase task data.

Area of Science:

• Behavioral Economics
• Biostatistics
• Addiction Science

Background:

• The Cigarette Purchase Task (CPT) assesses smoking reinforcement efficacy across prices.
• Current analysis relies on exponential demand equations, which struggle with zero consumption data.
• This limitation hinders accurate modeling of smoking behavior and cessation.

Purpose of the Study:

• To introduce a novel two-part mixed effects model for CPT data.
• To effectively model both zero and nonzero consumption values in behavioral economics.
• To retain the strengths of exponential demand equations while addressing zero consumption.

Main Methods:

• A two-part mixed effects model combining logistic regression (for zero consumption) and an exponential demand equation (for nonzero consumption).
• Application of the model to baseline data from 1,217 college students in a tobacco cessation trial.
• Monte Carlo simulations to compare model fit with existing methods.

Main Results:

• The proposed model demonstrated superior fit compared to existing models in simulations.
• The logistic regression component yields a "derived breakpoint" index.
• This index predicts the price threshold for likely smoking abstinence.

Conclusions:

• The proposed two-part mixed effects model offers an improved analytical approach for CPT data.
• It accurately models zero consumption while preserving the exponential demand framework.
• The derived breakpoint provides a valuable new metric for addiction research and cessation interventions, applicable beyond smoking to other substance use behaviors.