Analysing time-use composition as dependent variables in physical activity and sedentary behaviour research: different compositional data analysis approaches

Abstract

Recently, there has been a paradigm shift from considering physical activity and sedentary behaviour as “independent” risk factors of health to acknowledging their co-dependency and compositional nature. The focus is now on how these behaviours relate to each other rather than viewing them in isolation. Compositional data analysis (CoDA) is a methodology that has been developed specifically for compositional data and the number of publications using CoDA in physical activity and sedentary behaviour research has increased rapidly in the past years. Yet, only a small proportion of the published studies in physical activity and sedentary behaviour research have investigated the time-use composition as dependent variables. This could be related to challenges regarding the interpretation of the results and the lack of guidelines for deciding which statistical approach to use. Therefore, in this paper, four different approaches for analysing the time-use composition as dependent variables are presented and discussed. This paper advocates that the aim of research should guide how the dependent variable is defined and which data analysis approach is selected, and it encourages researchers to consider analysing time-use components as dependent variables in physical activity and sedentary behaviour research.