Abstract
Compositional data analysis (CoDA) has emerged as a method to analyse
complex physical activity (PA) data, including objective measurements from accelerometers. CoDA accounts for the dependence of PA behaviours in a single
statistical model, compared to standard techniques that incorrectly assume PA
behaviours are interdependent.
Although CoDA applied to cross-sectional data has been widely used, there
has been limited progress in extending this over multiple time points (i.e. longitudinally), especially when the outcome is compositional in nature. A review of the
current literature on CoDA applied to longitudinal data has been done as part of
this PhD project. It found that limited studies to date have analysed more than
two time points of accelerometry data using CoDA. In addition, methods used for
analysing two time points were limited and often could not explore trends within
individuals as simple statistical methods were used.
Previously, longitudinal analyses have been limited to univariate methods. A
compositional multivariate mixed effects model has been developed as part of
this PhD, to model multiple compositional outcomes in a single model. It is
more suitable for modelling the association between multiple compositional PA
behaviours and body mass index (BMI) standard deviation scores (SDS) over
time. This model will be presented and applied to accelerometry data from the
EarlyBird study.
EarlyBird is a longitudinal study of 347 healthy children, recruited from primary schools in Plymouth, UK. Annual observations of the children from age 5
up to and including age 16 were collected. The methods developed as part of
this PhD give a more informative picture of the association between PA and BMI
SDS over childhood and adolescence in boys and girls. Although these methods
have been demonstrated using PA data, they can be applied to any data that is
compositional, therefore making it applicable to many data sets.
complex physical activity (PA) data, including objective measurements from accelerometers. CoDA accounts for the dependence of PA behaviours in a single
statistical model, compared to standard techniques that incorrectly assume PA
behaviours are interdependent.
Although CoDA applied to cross-sectional data has been widely used, there
has been limited progress in extending this over multiple time points (i.e. longitudinally), especially when the outcome is compositional in nature. A review of the
current literature on CoDA applied to longitudinal data has been done as part of
this PhD project. It found that limited studies to date have analysed more than
two time points of accelerometry data using CoDA. In addition, methods used for
analysing two time points were limited and often could not explore trends within
individuals as simple statistical methods were used.
Previously, longitudinal analyses have been limited to univariate methods. A
compositional multivariate mixed effects model has been developed as part of
this PhD, to model multiple compositional outcomes in a single model. It is
more suitable for modelling the association between multiple compositional PA
behaviours and body mass index (BMI) standard deviation scores (SDS) over
time. This model will be presented and applied to accelerometry data from the
EarlyBird study.
EarlyBird is a longitudinal study of 347 healthy children, recruited from primary schools in Plymouth, UK. Annual observations of the children from age 5
up to and including age 16 were collected. The methods developed as part of
this PhD give a more informative picture of the association between PA and BMI
SDS over childhood and adolescence in boys and girls. Although these methods
have been demonstrated using PA data, they can be applied to any data that is
compositional, therefore making it applicable to many data sets.
| Original language | English |
|---|---|
| Publication status | Accepted/In press - 2024 |
| Event | The 10th International Workshop on Compositional Data Analysis - Girona, Spain Duration: 4 Jun 2024 → 7 Jun 2024 |
Conference
| Conference | The 10th International Workshop on Compositional Data Analysis |
|---|---|
| Country/Territory | Spain |
| City | Girona |
| Period | 4/06/24 → 7/06/24 |
ASJC Scopus subject areas
- Statistics and Probability
- Multidisciplinary
- Pediatrics, Perinatology and Child Health
Keywords
- accelerometry
- compositional data analysis
- physical activity
- statistics
- longitudinal
- cohort study
- multivariate