Skip to main navigation Skip to main content
The University of Southampton
Southampton Statistical Sciences Research Institute

Modelling health outcomes as a function of high dimensional physical activity profiles Seminar

Date:
24 November 2011
Venue:
Building 54 Room 10037

For more information regarding this seminar, please email Mrs Jane Revell at j.revell@southampton.ac.uk .

Event details

Statistics research seminar

Abstract
We develop a new statistical modelling approach for investigating the relationship between fat mass and physical activity measured by accelerometer. The key feature of the model is that it utilises the full profile of measured physical activity, rather than traditionally used scalar summary measures such as average daily moderate to vigorous physical activity (MVPA). In order to compare accelerometer profiles between individuals and to reduce the high dimension of the profiles a functional summary of the profiles is used. We compare different types of functional summaries including the quantile function, histogram and density function in terms of simplicity, suitability and interpretability. The histogram is the most useful functional summary due to its simplicity and ease of interpretation. The model used is a generalised regression of scalars on functions (Ramsay and Silverman, 2005) where an approximation to the integral of the product between the functional summary and a weight function enters the model as a predictor term. The estimated weight function is positive in the range of the accelerometer profile which has a positive contribution to fat mass and negative in the range of accelerometer counts with a negative contribution to fat mass. The estimated weight function changes from positive to negative at the currently used cut-point for moderate to vigorous activity. Hence our results independently confirm the MVPA cut-point currently used in ALSPAC. The model results are robust to changes in the functional summary, such as changing the number of bins of the histogram or using different transformations of accelerometer counts. We also find that changes in the data processing protocol make very little difference to the results. The results indicate that moderate and vigorous activity has a negative contribution towards fat mass and that sedentary and the lower end of light activity has a positive contribution towards fat mass. Ramsay, J.O. and B.W. Silverman, Functional Data Analysis, Springer, New York, 2005.

Speaker information

Dr Nicole Augustin , University of Bath. Senior Lecturer

Privacy Settings