Interval-censored failure time data arise when the time to a failure event is not exactly observed but known or observed only to fall within some time interval. Such data are collected in many fields, including epidemiology, medicine, demographics, economics, finance, psychology and social sciences. This research will propose new cost-effective sampling designs with interval-censored failure time data, in response to the needs of cost reduction and improved power for detecting the exposure-failure-time relationship in epidemiological and biomedical studies. The proposed designs coupled with robust and efficient statistical methods will enable study investigators to collect more informative samples and produce more precise statistical inference at a fixed budget, and they will be especially useful when the failure rate is low and the exposure variable is expensive or difficult to obtain. The graduate student support will be used for theoretical investigation of the proposed methods as well as conducting numerical studies and manuscript preparation.
The new designs proposed in this research are generally two-phase biased-sampling schemes where one observes the expensive exposure at phase II with a probability depending on the phase I information collected on failure times, cheap covariates, or auxiliary variables. All proposed designs will produce biased-sampled interval-censored data with missing covariates. No methods available in the literature can appropriately handle such data. This research will develop robust and efficient semiparametric likelihood-based methods that properly account for the proposed sampling schemes. Nonparametric techniques such as sieve and kernel estimation will be employed. New arguments based on modern empirical process theory will be developed for theoretical investigation of the proposed methods. Efficient computational algorithms and user-friendly software that implement the proposed designs and methods will be created and disseminated. Successful completion of the proposed research will have a significant impact on how future cost-effective epidemiological and biomedical studies be conducted and how data from these studies be efficiently analyzed. This research will also have potential impact on how to perform cost-effective studies and data analyses in other fields that produce interval-censored failure time data, such as demographics, economics, finance, psychology and social sciences.
This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.
