Study Design for Longitudinal and High Dimensional Measures.

Abstract

Study design is the foundation of successful clinical or epidemiological studies. Ever since the seminal work of Fisher (1935), research in this area has blossomed and many innovative concepts and approaches have been developed. Despite extensive literature on study design, new challenges for study design continue to emerge as innovative technologies push the limits of what can be investigated with a clinical or epidemiological study. For instance, tools for ecological momentary assessment of behaviors or biological markers, or high throughput experiment devices such as microarrays open the opportunity to measure complex biological processes over time, or the expression levels of millions of genetics or proteomics biomarkers simultaneously. In this dissertation, we develop novel design methodologies for studies employing these new data collection techniques, namely: 1) studies involving repeated measures of nonlinear profiles in biomarker studies with the objective of estimating features of the profile; 2) studies involving data with underlying functional response with the objective of capturing the mean profile and between subject variability; 3) studies involving high dimensional genetics and proteomics data with the objective of constructing classifiers with high probability of correct classification. Correspondingly, our research is motivated by three practical applications: 1) salivary cortisol studies for investigating the association between cardiovascular disease and stress; 2) urinary progesterone studies for reproductive health; 3) studies involving high dimensional genetics and proteomics data with the objective of constructing classifiers with high probability of correct classification. This dissertation contributes novel study design methodologies for studies that involve related but distinct data structures. We demonstrate the use of the methods with various examples to enhance the potential of their used across a variety of settings. The new design methodology will thus enable investigators to better evaluate the feasibility and cost-efficiency of the study in the planning stage and ultimately improve the chance of success of studies involving longitudinal and high dimensional data.