The broad, long-term objectives of this research are the developments of simple and useful statistical methods for the design and analysis of clinical and epidemiologic cancer studies with incomplete observations.
The specific aims i nclude (1) investigation of semi-parametric regression methods for assessing the effects of covariates (e.g., cancer therapy and patient characteristics) on medical cost and quality-adjusted lifetime based on incomplete follow-up data, (2) construction of non- and semi-parametric methods for the joint analysis of incomplete repeated measures (e.g., serial quality-of-life measures) and censored failure times (e.g., times to cancer recurrence/death) from longitudinal cancer studies, and (3) exploration of efficient methods of design and analysis for two-phase survival studies (e.g., case-cohort studies, sample surveys and covariate measurement error problems). The proposed statistical models and inference procedures are built from but extend significantly the current knowledge about the analysis of censored failure time data and incomplete repeated measures. These models are highly flexible and versatile in that they do not require specifying the distributional form of any random variable or the dependence structure between any two related outcome measures. The asymptotic properties of the proposed estimators and test statistics will be investigated rigorously with the use of counting-process martingale theory, modern empirical process theory and other probability tools. Their operating characteristics in practical settings will be evaluated extensively through computer simulations. The usefulness of the proposed methods will be illustrated with real cancer studies. The research results will be disseminated to practicing statisticians and medical investigators via publications, lectures and software distributions.
Showing the most recent 10 out of 104 publications