Date of Award
Doctor of Philosophy
E. Olusegun George
Dale Bowman Armstrong
Time to event is a commonly used endpoint in epidemiological and disease prevention trials in order to study the relationship between risk factors and the endpoints. Case-cohort design that consists of a sub-cohort randomly sampled from full cohort, and all subjects with event is often applied in studies where the disease is rare and the cost of collecting the event information is high. with the non-rare events, a generalized case-cohort design is advocated in which a subset of events instead of all events is sampled. Cai and Zeng have proposed the general log-rank tests and the corresponding sample size/power formulas to compare the hazard rates between two groups under the case-cohort and the generalized case-cohort designs, respectively. However, in many practical situations, the population is not homogenous and stratification is considered. While prequalification is increasingly commonly used in large cohorts, the stratified log-rank tests and the sample size and power estimation techniques have not been available even though these issues are critical to the study design. this dissertation is devoted to consider these issues and fulfill the availability. In addition to the development of the stratified general log-rank tests and the sample size/power formulae for both the stratified case-cohort design and the stratified generalized case-cohort design, simulation studies are to be conducted to examine the performance of the tests. Furthermore, optimal, proportional, and balanced sampling strategies are to be explored and recommendations are to be made. Two real epidemiological studies are to be presented to illustrate the sample size calculation under these sampling strategies.
dissertation or thesis originally submitted to the local University of Memphis Electronic Theses & dissertation (ETD) Repository.
Hu, Wenrong, "Sample Size/power Calculation for Stratified Case-cohort Design and Generalized Stratified Case-cohort Design" (2013). Electronic Theses and Dissertations. 835.