Statistical significance of the day-to-day variation (within-subject) in physical activity measures due to differences in life-space area reached was tested; within each individual, days on which the same life-space area was reached were clustered. The variation in physical activity (dependent variable with log link transformation) according to the life-space area (betweensubject factor) reached on that respective day was tested using General Estimating Equation (GEE) models with unstructured working correlation matrix. Day of assessment was included as within-subject factor. All participants moving through at least two different life-space areas during the assessment week (N = 150) were included in the analyses. GEE modeling was conducted with and without multivariate imputation by chained equations (MICE) procedure [24] in SPSS (version 20.0; IBM, Armonk, NY, USA). The results of these analyses were similar; only the imputed data are reported. Analyses estimating step counts, moderate and low activity time, and sedentary time were adjusted for accelerometer wear time. Subsequently, the analyses were further adjusted for age, sex, and potential confounders. The life-space area “neighborhood” was used as reference group in the analyses as this area most likely encompasses activities that contribute to the accumulation of physical activity. Of the 150 participants included in these analyses, 59% visited this area at least once during the assessment week. Supplementary analyses were conducted using the life-space area “inside only” as the reference group; 20% of the included participants restricted their life-space area at least once to the home. Finally, for each individual, the physical activity scores were determined for the life-space areas reached
(means in case of recurrence) and used to calculate the median score and interquartile range of the physical activity in the different life-space areas. Statistical significance was set at P < .05.