Teun Remmers

General discussion | 191 compositional data analyses may reveal interesting co-dependencies between these behaviors that in turn may have different consequences on children's health (66).  Complex conceptual models involving mediation and moderation provides researchers with valuable deeper insight into how determinants influence PA. In addition to the relatively isolated approach of moderation as described in this thesis, more flexible approaches such as structural equation modelling or agent- based modelling have the potential of simultaneously investigating multiple moderation or mediation mechanisms (88, 91). We recommend embedding these analyses in well-conceived theoretical frameworks.  Researchers should take into account that combined accelerometer and GPS methodologies may lead to decreased compliance in wear time, especially with repeated measurements. Also, due to potential concerns about privacy and data- storage, selection bias may occur. Therefore, future studies are encouraged to develop and communicate clear protocols regarding data-storage and analyses, as well as consider oversampling the population of interest.  The increased acknowledgement of the importance of regular PA has led to recent advances in consumer-level PA monitors. These monitors tend to be more user-friendly, relatively cheap, allow users to upload and review their data in online applications, and can be worn on the hip or wrist (124). As some consumer-level monitors tend to perform well in capturing daily accumulated steps compared to more advanced research-level monitors, this may offer exciting possibilities to analyze participant's PA in real-time (124-126). Although these innovations may be inspiring, there are still numerous additional challenges to overcome before consumer-level monitors can be used for research purposes; for example, these include improving concurrent-validity of especially MVPA classifications, data ownership and privacy, blinding participants for individual feedback in observational studies, and identification of cycling and motorized transport.  Driven by fast developing technological innovations, researchers are strongly encouraged to continue their efforts into the feasibility, reliability and validity of new research-level monitors to quantify PA, sleep, and sedentary behaviors. Of special interest are smaller, more comfortable monitors, with enhanced battery- life and data-storage capacities. While combining data from multiple sensors, these devices may be able to recognize daily composition (i.e. sleep, sedentary behaviors, and PA) and specific PA contexts (e.g. bicycling, sports participation). As these new devices may cause less burden for participants, study designs with considerably longer monitoring periods may be feasible. With these longer monitoring periods, researchers may be able to, for example, investigate compensation-mechanisms across longer time periods (64).