Robust Compositional Analysis of Physical Activity and Sedentary Behaviour Data.

Area of Science:

• Public health research
• Biostatistics
• Adolescent health

Background:

• Increasing recognition of compositional data methodology's utility in public health.
• Traditional statistical methods have been predominantly used, potentially overlooking data nuances.
• Need for advanced statistical approaches to model complex behaviors like physical activity and sedentary time.

Purpose of the Study:

• To demonstrate the application and benefits of robust statistics in compositional data analysis for public health.
• To investigate the inter-relationships between different physical activity intensities and their association with age in adolescents.
• To explore the links between adolescents' physical activity and sedentary behavior structure and their obesity status.

Main Methods:

• Utilized compositional data analysis, including compositional covariates, response, and regression between parts.
• Employed robust statistical methods as counterparts to classical regression to mitigate outlier influence.
• Analyzed movement behavior data from Czech adolescents, focusing on physical activity intensities and sedentary behavior.

Main Results:

• Identified a pattern where extensive sedentary behavior (SB) correlated with higher light-intensity physical activity (PA).
• Vigorous PA ratios were identified as a potential source of aberrant observations in the data.
• Aging in adolescents was associated with increased SB and vigorous PA, decreasing light and moderate PA.

Conclusions:

• Robust statistical methods offer more stable estimates when dealing with potential outliers in compositional data.
• Findings suggest that substituting sedentary time with vigorous physical activity could be an effective strategy against adolescent obesity.
• Compositional data analysis, particularly robust approaches, provides valuable insights into adolescent movement behaviors and health outcomes.