This project will address methodology development needs that are important in chronic disease population research. These include non- parametric estimator of multivariate survivor functions as relevant to disease prevention trials with multiple clinical outcomes and to family studies in genetic epidemiology. The closely related auxiliary data problem under which one uses data on multiple short-term response variables to strengthen the analysis of a primary clinical outcome will also be considered. Both estimators based on Peano series representations of the survivor function and likelihood-based estimators will be considered. Measurement error in key exposure and confounding variables can be a critical factor in determining observational study reliability. A more flexible than usual measurement model will be studied for application to such difficult to measurement exposures as nutritional and physical activity factors. Regression calibration estimators and empirical score function estimators of relative risk parameters will be studied based on data from this new measurement model in conjunction with an objective measures (biomarker) subsample.
A third aim combines aspects of multi-variate failure time analysis and co-variate measurement modeling to develop new estimators of pairwise dependency in ages at disease onset of family members in the context of genetic epidemiology studies with measurement error in environmental factors. This project will also continue our studies of the role of various population study designs in the chronic disease research agenda as a function of the biases, measurement issues, and other limitations of ecologic studies, analytic epidemiologic studies, small intervention trials and full-scale intervention trials using the Women's Health Initiative database as a principal motivating setting.
Showing the most recent 10 out of 319 publications
