A Causal Approach to Functional Mediation Analysis with Application to a Smoking Cessation Intervention

Area of Science:

• Biostatistics
• Psychology
• Digital Health

Background:

• Mobile and wearable devices facilitate intensive longitudinal data collection.
• Understanding mediating processes is crucial for evaluating treatment effectiveness.

Purpose of the Study:

• To define and estimate causal direct and indirect effects in longitudinal data.
• To develop a method for assessing mediation in randomized treatments using intensively measured variables.
• To provide an R package for applying the proposed mediation analysis technique.

Main Methods:

• Utilizing potential outcomes framework to define causal effects.
• Estimating indirect effects on mean difference and odds ratio scales.
• Employing nonparametric trajectory analysis for longitudinal mediator.
• Demonstrating a bootstrap test for indirect effect significance via simulation.

Main Results:

• The study successfully defines and estimates causal indirect effects.
• A simulation study validates the bootstrap test for indirect effects.
• An empirical example demonstrates application in smoking cessation, linking craving patterns to abstinence.

Conclusions:

• The proposed method effectively estimates indirect effects in intensive longitudinal data.
• The R package 'funmediation' facilitates the application of this technique.
• Future research can extend this to multiple mediators and complex intervention designs.