Teun Remmers

Investigating PA patterns in the transition from primary to secondary school | 157 Statistical analyses First, descriptive statistics were performed in order to present children's general PA patterns at primary and secondary school. After describing the development of total PA in the transition between primary and secondary school, we calculated the median number of minutes for the time that children spent in context-specific LPA and MVPA activities, both in primary school and secondary school (45). Subsequently, multi-variate analyses were performed to test whether context-specific changes significantly increased or decreased during the transition from primary to secondary school. As context-specific PA-variables showed significant differences from normality, we decided to present median parameters in our univariate descriptive analyses, and log-transformed transition coefficients in our multivariate explanatory analyses. In subsequent multivariate models, days were used as the unit of analyses and the main variables of interest was the index-variable that represented either baseline- or follow-up measurement. All multivariate analyses were performed using multilevel linear mixed models, with a repeated term for days within children and a random term for children within schools. Outcome variables were minutes of sedentary time (ST), LPA and MVPA (specified for each time-segment and context). First, we adjusted our multilevel models for context-specific wear time, gender, and age. In addition, based on registries from local weather stations, we extracted hourly meteorological data for each day of PA measurement. We computed daily averages of meteorological data between 6 AM and 8 PM (as only during these times it would potentially affect children's behaviours), and we adjusted for these meteorological variables in the second set of analyses. Model fit and normality of residuals of these models were inspected to verify its fitting capabilities. Some contexts consisted of very few data-points (e.g. during school time at shopping centres). In order to refrain from presenting underpowered and under-fitted results, 1) non-significant covariates ( p > 0.05) were deleted from models, and 2) contexts with a median of less than 5 minutes of data and performed for less than 50 days were not presented in subsequent analyses. We also tested for moderation-mechanisms in our models. First, we investigated moderation of gender by computing interaction terms (i.e. gender*transition coefficient) and inspecting results for boys and girls separately. Likewise, we also investigated moderation of the difference in distance from home to primary school versus home to secondary school (hereafter referred as difference in home-school distance) on the development of PA patterns in the transition from primary to secondary school. In subsequent analyses, stratum specific associations were performed after computing a median split of home-school differences (median = 2215 meters). Statistical analyses were performed using SPSS 21.0 for Windows (IBM SPSS Inc., Armonk, NY), and p < 0.05 indicated statistical significance.