Modeling the Measurement Error Structure of Physical Activity Data

• Activity Number: 487
• Type: Invited
• Date/Time: Wednesday, August 3, 2011 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Sarah Nusser*+
• Companies: Iowa State University
• Keywords: measurement error ; physical activity

Abstract

Activity data take many forms (e.g., energy expenditure, metabolic equivalents (METS), time spent sedentary activities) and are often collected in epidemiological studies. Although most physical activity data collection methods generate data with substantive measurement error, statistical error adjustments are not yet commonly applied to activity data when making inferences. As with dietary intake, the concept of "usual" activity is of interest when making inferences about long-run average behaviors for individuals and about the relationship of activity to chronic health outcomes. Using data from a survey in which respondents provided duplicate observations of concurrent activity recall and activity monitor data, we will present models to estimate error characteristics of recall and sensor data measurements. We will also use a measurement error framework to estimate the distribution of usual activity for subpopulations to characterize behaviors at the population level.