Adjusting for Measurement Error and Nonresponse in Physical Activity Surveys: A Simulation Study

Abstract

Abstract Adult Americans are encouraged to engage in at least 150 minutes of moderate to vigorous physical activity (MVPA) each week. National surveys that collect physical activity data to assess whether or not adults adhere to this guideline use self-report questionnaires that are prone to measurement error and nonresponse. Studies have examined the individual effects of each of these error sources on estimators of physical activity, but little is known about the consequences of not adjusting for both error sources. We conducted a simulation study to determine how estimators of adherence to the guideline for adults to engage in 150 minutes of MVPA each week respond to different magnitudes of measurement and nonresponse errors in self-reported physical activity survey data. Estimators that adjust for both measurement and nonresponse errors provide the least amount of bias regardless of the magnitudes of measurement error and nonresponse. In some scenarios, the naïve estimator, which does not adjust for either error source, results in less bias than estimators that adjust for only one error source. To avoid biased physical activity estimates using data collected from self-report questionnaires, researchers should adjust for both measurement error and nonresponse.