2016-Biostatistics: Survival and Current Status Data

Biostatistics: Survival and Current Status Data 
Chair: Cindy Feng (University of Saskatchewan) 

LONGLONG HUANG, University of Calgary
A Group Bridge Approach for Component Selection in Nonparametric Accelerated Failure Time Additive Regression Model  [PDF]
We propose a nonparametric accelerated failure time additive regression model whose covariates have nonparametric effects on the survival time. The proposed model can be fitted to high-dimensional censored data. B-splines are used to approximate the nonparametric components. A group bridge penalized variable selection approach based on the inverse probability-of-censoring weighted least-squares is developed to select nonparametric components. Simulation studies indicate that the proposed method is able to distinguish the nonzero components from the zero components and estimate the nonzero components simultaneously even with relatively high censoring rates. Real data analysis illustrates the application of the proposed method to survival data. 
SAIMA KHOSA, University of Saskatchewan
A Generalization of Log-logistic Distribution with Application in Survival Analysis  [PDF]
The log-logistic distribution has wide applications in analyzing survival data. The model is closed under both multiplication of failure time and proportionality of odds. However, it is not a proportional hazard (PH) model. In survival analysis, PH models play a pivotal role in many applications. So in this study, we propose a generalization of the log-logistic distribution that falls into the PH family. A comparison between the generalized log-logistic and the Cox PH model reveals that the generalized log-logistic PH model performs reasonably well in analyzing different types of time-to-event data. 
SHANSHAN LU, University of Calgary
Exploring the Varying Covariate Effects in Partially Linear Proportional Odds Models with Current Status Data  [PDF]
We consider a partially linear proportional odds model with current status data. This model enables one to examine the extent to which some covariates interact nonlinearly with an exposure variable, while other covariates present linear effects. B-spline approach and sieve maximum likelihood estimation method are used to get an integrated estimate for the linear coefficients, the varying-coefficient functions and the baseline function. The proposed parameter estimators are proved to be consistent and asymptotically normal, and the estimators for the nonparametric functions achieve the optimal rate of convergence. Simulation studies and a real data analysis are used for assessment and illustration. 
YUAN DONG, University of Calgary
Efficient Estimation of Varying-Coefficient Partially Linear Proportional Hazards Models with Current Status Data  [PDF]
We consider a semiparametric varying-coefficient proportional hazards model with current status data. This mode enables one to assess possibly nonlinear effect of a certain covariate on the hazard rate. We apply B-splines to approximate both the unknown baseline hazard function and the varying-coefficient function. Sieve maximum likelihood estimation method is used for estimation. The rate of convergence of the estimators for the two unknown smooth functions is obtained and the estimator for the unknown parameter is shown to be asymptotically efficient and normally distributed. Simulation studies and data analysis are conducted to examine the finite-sample properties of the proposed estimators. 
A Comparison of the Penalized Estimation Methods in the Cox Model for Right-censored Length-biased Data  [PDF]
Length-biased data commonly arises in epidemiological studies when prevalent sampling is used for recruiting cohort subjects. Several estimation methods have been proposed for parameters in different survival models in the presence of length-bias in the last decade. In this talk, we will focus on variable selection and discuss how to penalize those methods with different penalty functions. We will compare the resulting estimators for the parameters in the Cox model in terms of their efficiency and employ the selected method to analyze a real right-censored length-biased data. 
DONGDONG LI, Simon Fraser University
Modelling and Analysis of Bivariate Event-Times with Informative Censoring  [PDF]
In attempt to evaluate risk of cancer patient to cardiovascular disease, we consider analysis of bivariate event-times with informative censoring. Our particular attention is to the inherent challenges to modelling association between the two event-times and their dependence with the censoring time. We explore various models and develop inference procedures under proposed models. The research is motivated and will be illustrated using records of cardiovascular disease related hospitalization from a cohort of breast cancer patients together with their cancer registry information. This is joint work with Professor Joan Hu.