Department of Statistics Seminar
North Carolina State University

presents

Xiao Song, Yao Huang, Bingming Yi

North Carolina State University

A Sampler of Graduate students

ABSTRACT

Three graduate students of the Department of Statistics at NCSU will present brief overviews of their current research.

• Xiao Song- An estimator for the proportional hazards model with multiple longitudinal covariates measured with error
  In many longitudinal studies, it is of interest to characterize the relationship between a time-to-event (e.g. survival) and several time-dependent and time-independent covariates. Time-dependent covariates are generally observed intermittently and with error. In the case of a single time-dependent covariate, a popular approach is to assume a joint model for the longitudinal data and survival, where the time-dependent covariate follows a linear mixed effects model and the hazard of failure depends on random effects and time-independent covariates via a proportional hazards relationship. Regression calibration and likelihood-based methods have been advocated to implement these models. However, in the case of multiple time-dependent covariates these methods become prohibitive. Recently, Tsiatis and Davidian (2001) have proposed an approach in the case of a single time-dependent covariate that is easily implemented and moreover does not require an assumption on the distribution of the longitudinal random effects. Unlike other approaches, this technique may be extended to the case of multiple, possibly correlated, time-dependent covariates, as we demonstrate. We illustrate the approach via simulation and by application to data from an HIV clinical trial.

• Yao Huang - A Conditional Nonparametric Approach to Estimating Survival Functions
  A Conditional Nonparametric Approach to Estimating Survival Functions Abstract: Here we propose a new conditional nonparametric approach to estimating the survival distribution of censored data. Our method uses a data augmentation step by sampling from a nonparametric conditional distribution of unobserved times given the censored observations. The proposed estimator is obtained iteratively as a solution to estimating equations. A proof of consistency of our estimator is given under some mild conditions. We compare the performance of our estimator to the popular Kaplan-Meier estimator using extensive simulations.

• Bingming Yi - Latent Class Regression Analysis on the Potency of Chemical Compounds and Comparison to Cell Based Analysis and Recursive Partitioning
  Drug discovery is dependent on finding a very small number of biologically active or potent compounds among billions of compounds stored in chemical libraries. Quantitative structure activity relationships suggest that potency of a compound is highly related to that compound's chemical makeup or structure. As such, statistical models that predict potency based on chemical structure can be very cost-effective since it would eliminate the need to individually test all the billions of compounds included in the library. Results from several cell based analysis methods, Helixtree recursive partitioning and latent class regression are compared. A real dataset from National Cancer Institute is used.

Friday, September, 28, 2001

3:35 - 4:35 pm

206 Cox Hall

Refreshments will be served on the second floor of Dabney Hall (left of Room 222) at 3:00 pm.