Biostatistics Seminar Series: Ying Ding, PhD

Title: Copula-Based Semiparametric Sieve Models for Bivariate Interval-Censored Data Abstract: This research is motivated by discovering genetic causes for the progression of a bilateral eye disease, Age-related Macular Degeneration (AMD), where the primary outcomes, progression times to late AMD, are bivariate and interval-censored. We develop a flexible copula-based semiparametric approach for modeling and testing bivariate interval-censored data. Specifically, the joint likelihood is modeled through a two-parameter Archimedean copula, which can flexibly characterize the association structure between two margins. The marginal distributions are modeled through a semiparametric transformation model using sieves, with the proportional hazards model being a special case. We propose a two-step maximum likelihood estimation procedure and develop a computationally efficient score test, which is suitable for large scale testing as we consider here. We establish strong consistency and asymptotic normality for the proposed estimator.  Extensive simulations are conducted to evaluate the performance of the proposed method in finite samples. Finally, we apply our method on a genome-wide analysis of AMD progression, to identify susceptible risk variants for the disease progression.