Statistical challenges in observational cohort studies

For over a century observational cohort studies have been used to study determinants of health and disease. Within a sample from the population, we can determine the relation between health outcomes (e.g. death) and a broad range of factors as genetic markers, environmental exposures, and lifestyle determinants.
Observational cohort studies are effective tools to investigate determinants of health and disease. Importantly, in this type of study the researcher only observes and does not assign interventions or characteristics to the individuals in the sample drawn from the population of interest. In this thesis, four subjects regarding the design and execution of cohort studies have been investigated. The first part is dedicated to problems that arise when our goal is to recruit a specific sample from a finite population. In the second part we propose methods to analyze of data derived from record linkage. In the third part we consider joint models with a large number of (recurrent) events and markers. Finally, in the fourth part, we analyzed childhood growth data from a large prospective cohort study.