Powerful Actigraphy Data Through Functional Representation

• Activity Number: 154
• Type: Invited
• Date/Time: Monday, July 30, 2012 : 10:30 AM to 12:20 PM
• Sponsor: IMS
• Author(s): Jimin Ding*+ and Juergen Symanzik and Abbass Sharif and Jia Wang and Stephen Duntley and William Shannon
• Companies: Washington University in St. Louis and Utah State University and Utah State University and Washington University in St. Louis and Washington University in St. Louis and Washington University School of Medicine
• Keywords: Actigraphy ; functional data analysis ; functional linear model ; functional principal components analysis

Abstract

An actigraph is a watch-like device that is usually attached to the wrist or the leg, and contains accelerometers to measure movement in the form of an activity count every minute or every few seconds. As an emerging clinical technology, actigraphy data is often collected over several days in trials, to evaluate the sleep, daytime activity, and circadian activity rhythms of each participant. In this paper, we view the measured daily activity counts for each person as a function of time, and analyze them using functional data analysis (FDA). Functional linear models are applied to the obtained curves. We employ principal components analysis to decompose covariances to fit the model.