Modeling daily and weekly moderate and vigorous physical activity using zero-inflated mixture Poisson distribution

Area of Science:

• Epidemiology
• Biostatistics
• Physical Activity Measurement

Background:

• Accelerometers provide objective physical activity data in large studies.
• Traditional models struggle with zero-inflated daily activity data and weekly patterns.
• Existing methods limit comprehensive analysis of physical activity variations.

Purpose of the Study:

• To propose a novel statistical approach for modeling daily and weekly physical activity.
• To address limitations of zero-inflated models in capturing activity heterogeneity.
• To enable simultaneous analysis of inactivity and activity intensity.

Main Methods:

• Utilized a zero-inflated Poisson mixture distribution for daily and weekly physical activity.
• Employed a joint random effects model to account for participant heterogeneity.
• Applied Gaussian quadrature technique for maximum likelihood estimation via the GLMMadaptive R package.

Main Results:

• The proposed method effectively models both daily and weekly physical activity within a unified framework.
• Simulation studies demonstrated the method's performance in capturing activity variations.
• Application to the Hispanic Community Health Study/Study of Latinos revealed associations between physical activity and BMI groups, and weekday/weekend differences.

Conclusions:

• The zero-inflated Poisson mixture model offers a flexible and robust approach for analyzing accelerometer-derived physical activity data.
• This method enhances the understanding of physical activity patterns and their health implications.
• The approach is valuable for epidemiological research and personalized health assessments.