Compositional data analysis as an alternative paradigm for nutritional studies.

Area of Science:

• Nutritional Science
• Biostatistics
• Epidemiology

Background:

• Dietary data is compositional, meaning parts are relative.
• Standard statistical methods can yield spurious results with compositional data.
• A compositional data approach is proposed for nutritional analysis.

Purpose of the Study:

• To apply a compositional data approach to nutritional data analysis.
• To examine isocaloric associations between dietary balances and waist circumference.
• To provide a holistic overview of dietary component roles.

Main Methods:

• Utilized isometric log-ratio (ilr) transformation on macronutrient, vitamin, and mineral compositions.
• Applied sequential binary partitions to define specific dietary balances.
• Included ilr-transformed variables as covariates in linear regression models.

Main Results:

• Total energy intake was positively associated with waist circumference (WC).
• Fiber intake showed a negative association with WC.
• Relative increases in starches and niacin, and levels of Vitamin C and folate, were associated with WC.

Conclusions:

• The compositional data approach enhances interpretation of relative dietary component roles.
• This method provides valuable insights into diet-waist circumference relationships.