Parsimonious Covariate Selection for Interval Censored Data

TitleParsimonious Covariate Selection for Interval Censored Data
Publication TypeThesis
Year of Publication2020
AuthorsCui, Y
Academic DepartmentBiostatistics
DegreeDoctor of Philosophy
UniversityState University of New York at Albany
CityAlbany, NY
ISBN Number9781658404419
KeywordsParsimonious covariate selection, Statistics

Interval censored outcomes widely arise in many clinical trials and observational studies. In many cases, subjects are only followed-up periodically. As a result, the event of interest is known only to occur within a certain interval. We provided a method to select the parsimonious set of covariates associated with the interval censored outcome. First, the iterative sure independence screening (ISIS) method was applied to all interval censored time points across subjects to simultaneously select a set of potentially important covariates; then multiple testing approaches were used to improve the selection accuracy through refining the selection criteria, i.e. determining a refined common cutoff value. We compared the improvement of selection accuracy by using both familywise error rate (FWER)and generalized FWER (gFWER) methods. Our method shows good performance in simultaneously in selecting non-zero effects and deselecting zero-effects, respectively.

Citation Key11102