05/20/2009 — The Proportional Odds Model in the Sequential Monitoring of Clinical Trials

Ken Wu

In this talk, I will introduce a work in progress regarding the monitoring of clinical trials with the proportional odds model. Unlike the proportional hazards model, the hazards converge under the proportional odds model, and we believe this will more closely model treatment arms whose effects wane over time. I will begin by introducing sequential monitoring and some of the models and statistics currently used, and I will then describe the settings in which the proportional odds would be preferred. After discussing the origins and interpretation of the proportional odds model, I will talk about its estimation and some semiparametric efficiency issues that we have encountered. Finally, I will discuss the extension of this estimation to multiple interim analyses and list what questions remain.

05/13/09: Time-dependent Predictive Accuracy: Extending Binary Classification Accuracy Methods for Censored Survvival Data

Paramita Saha

I will be talking about my dissertation. In this dissertation, we characterize the prognostic value of a scalar score or a marker by extending standard binary classification accuracy summaries like sensitivity or True Positive (TP),  specificity or 1 – False Positive (FP), Receiver Operating Characteristic (ROC) curve and Area Under the ROC Curve (AUC) for censored survival data. In my dissertation, I introduce novel statistical methodology to solve some of these problems. First, we propose time-dependent accuracy measures for a marker when we have censored survival times and competing risks. We extend time-dependent definitions of TP and FP to incorporate causes of failure for competing risks outcomes. The proposed methods extend the time-dependent predictive accuracy measures of Heagerty et al. (2000), and Heagerty and Zheng (2005). Next, we propose a direct, non-parametric estimator of the time-dependent AUC curve, and show that the proposed estimator performs comparably or better than the semi-parametric AUC curve estimator proposed by Heagerty and Zheng (2005).  The proposed method extends non-parametric AUC estimates for the binary data and we establish asymptotic properties. An overall measure of concordance is also proposed. Time-dependnet markers can also be accommodated in the estimation to capture the evolving nature of the marker. Finally, we introduce a time-averaged ROc curve to summarize the predictive accuracy of a marker accrued over time. We also introduce methods for comparison of markers via this summary ROC curve and demonstrate that this approach may be used to optimize screening schedules for diseases like breast cancer.

05/06/2009 — Recent thoughts on survival analysis

Gary Chan

I will present some recent research interests on survival analysis methodologies in three broad areas. (1) Empirical Bayes and prediction of survival probabilities. (2) Prevalent cohort covariate bias and information. (3) Recurrent marker process with informative terminal event. The methods discussed would be slightly difficult in focus from the mainstream survival analysis, and I hope this will stimulate students to view survival analysis in a broader context.

04/29/2008: Panel Discussion – Finding a Thesis Adviser/Topic

This week will feature a panel discussion and question/answer session about the sometimes intimidating process of finding a thesis adviser/topic. The event will be students-only, and snacks will be provided; don’t miss out!

04/22/2009 — Missing Data in Genome Wide Association Studies

Matthew Bryan

The study of genetic association is a rapidly growing area of statistical research. Due to the massive amounts of data generated by these studies, many problems arise in analyzing such data such as multiple comparisons, computational issues, and missing data. This presentation will discuss portions of an ongoing research project on missing data in genetic association studies. The goal of the project is to study genetic association methods, understand how these methods account for missing data, and assess whether these methods can be improved by further adjustment for missing data. This presentation will focus on a method suggested by Timothy Thornton and Mary McPeek that uses a Quasi-Likelihood Score Test approach. The method incorporates information from subjects that are missing phenotype and genotype information. Further discussion will look into extending these methods to better account for missingness through multiple imputation.