Multivariate variable selection in N-of-1 observational studies via additive Bayesian networks

Area of Science:

• Biostatistics
• Behavioral Science
• Network Analysis

Background:

• N-of-1 study designs are valuable for understanding individual variability over time.
• Traditional statistical models may not fully capture complex observational relationships in N-of-1 data.
• There is a need for advanced methods to analyze longitudinal, single-subject data.

Purpose of the Study:

• To propose and validate a novel statistical method for analyzing N-of-1 observational data.
• To model data-driven associations between variables within an individual over time.
• To explore the relationship between stress and daily exercise engagement in a 12-month N-of-1 study.

Main Methods:

• An additive Bayesian network was developed using a generalized linear mixed-effects model for the local mean.
• The proposed method was validated through simulation studies.
• The approach was applied to a 12-month observational N-of-1 study on stress and exercise.

Main Results:

• The additive Bayesian network demonstrated improved performance in recovering underlying network structures compared to traditional methods.
• The method successfully modeled associations within the N-of-1 study.
• Statistically discernible population-level associations between stress and physical activity were found, but individual-level differences were highlighted.

Conclusions:

• Additive Bayesian networks offer a powerful and flexible approach for modeling complex relationships in N-of-1 observational studies.
• The findings underscore the importance of considering individual-level variations in the relationship between stress and physical activity.
• This novel method enhances the analysis of personalized health and behavior data.