Teun Remmers

188 | Chapter 10 participants participated in measurements. As no data were obtained from children and schools that refused to participate, it is not possible to investigate the magnitude of potential selection bias. In chapter 7, data was used from the Active Living study (96), which focused on the development of tailored school-centered PA interventions in deprived areas in the south of the Netherlands. The majority of included participants were of Dutch ethnicity (97). Chapter 4 used data from the Be Active Eat Right Study, of which 83% of the parents had mid-high education, and 91% of the children were of Dutch ethnic background (98). In general, the Netherlands is a supportive environment for active transport. This is for example because of its highly connected streets, flatland, general nearness of children's homes to their schools, and absence of motorized transport infrastructures between home and school. In chapters 8 and 9, data was collected from a relatively urbanized area, whereas data from chapter 7 also included more rural areas. As children in rural areas often have bigger school-home distances, results between these studies may not be easily comparable. Therefore, future researchers may investigate whether children in rural areas have deviant PA patterns compared to children living in more urbanized areas. The study presented in chapter 6 was based on data from the PHASE study, which was carried out in Melbourne, Australia. The meteorological circumstances (especially high temperatures) experienced in these regions may be not generalizable to North-European countries, as temperatures tend to be lower especially in spring and summer (99, 100). Collectively, the majority of results presented in this thesis were collected from children from relatively highly educated Dutch parents, and in (physical and social) environments supportive of active transport. Therefore, results presented in this thesis may not be generalizable to different samples with other environments, generally lower educational levels, or with a more diverse ethnic makeup. Although our results may not easily be generalized, the methodology used in especially chapters 7, 8 and 9 may provide researchers with examples on how to investigate time-specific and location-specific associations between environmental determinants and PA, also in other samples and countries. Effect Sizes In the majority of the studies presented in this thesis, effect sizes may be considered small, and highly dependent on the context. This is in line with comparable studies focusing on PA (102). For example, chapter 2 describes a study that showed that an increment of 6.5 minutes of MVPA over at least three days was associated with 0.03 lower BMI scores. Another example is from chapter 7, which describes a study on environmental determinants of afterschool PA in the home-school environment. Results showed an influence of specific attributes of the built environment, but these were highly context- dependent (i.e. afterschool leisure time PA, cycling and walking). These small effect sizes may for example be influenced by measurement errors in the assessment of the environment using GPS and GIS methodologies. Namely, GPS loggers in general tend to make errors in determining locations; whereas the specific device used in chapters 8 and 9 performs somewhat better in both static and free-living conditions (59, 101). Although