This continuing project aims to develop and evaluate improved statistical methods for the design and analysis of epidemiologic studies. There will be a substantial emphasis on methods development for genetic epidemiologic studies in the proposed grant period. A multistage design that incorporates aggregation, segregation, linkage and candidate gene analysis will be developed. Properties of such a design will be studied numerically using likelihood-based procedures. Estimating equation-based methods will be developed for data arising from such a design, including emphasis on both discrete/continuous outcomes and censored failure time outcomes. Frailty methods will also be developed for the failure time analysis of epidemiologic data on families and will be compared with the estimating equation methods. There will also be a continuing emphasis on traditional cohort and case- control studies of identified risk factors. This work will include the development and evaluation of methods for nonparametric relative risk estimation. It will also include an emphasis on covariate measurement error methods development, including the relaxation of classical normal theory-based measurement methods to allow the measurement error distribution to depend on selected study subject characteristics, and the further development of less parametric procedures for use with a validation sample, or a partial validation sample. Finally this project will continue the development of aggregate data (ecologic) study methods, heretofore conducted under Project 4. In particular, design and analysis aspects of a study procedure that combines disease rates from several -geographic areas with exposure and confounding factor data from randomly selected persons in each area, will continue to be developed, as will related aspects of a procedure involving one or more time series of disease rates. Collectively this work can be expected to generate new design and analysis procedures that will improve the efficiency and reliability of a range of important types of epidemiologic studies.
Showing the most recent 10 out of 319 publications
