Abstract:
|
Accelerometers are often used to objectively measure physical activity in epidemiological studies. In these studies, it is interested to investigate whether sedentary behavior (SB), light physical activity (LPA) and moderate and vigorous physical activity (MVPA) are independently related to health outcomes. However, these variables often have high collinearity. Several models have been used in the literature, but they often yield conflicting conclusions. To solve the puzzle, we consider a compositional data perspective and propose a nonparametric compositional data analysis approach. The proposed approach is flexible and robust, unless existing models that rely on restrictive assumptions. We applied the proposed approach to an ancillary study of the Women’s Health Initiative to study the association between physical activity and cardiometabolic biomarkers in 6,500 older women.
|