The University of Texas MD Anderson Cancer Center
Department of Biostatistics
My biostatistical research covers two areas: propensity score analysis and joint analysis of longitudinal and survival data.
Propensity score analysis is an important method to adjust for confounding, and it is widely used in observational studies, which constitute a large majority of published evidence on the effectiveness of medical treatments. My research on propensity score analysis focuses on improved methods for efficient estimation, accurate variance estimation and objective balance checking. I am also extending the ideas to the generalized propensity score analysis, which studies the causal effect of continuous exposure variables. Given the fundamental role that propensity score plays in the broad research area of causal inference, this methodology can be extended in innovative ways in a variety of causal applications.
Most longitudinal cohort studies collect both longitudinal data and time to event data. These two types of data often need to be analyzed jointly to reveal the relationships between them. My research targets two aspects of the problem. First, when the trajectory of the longitudinal data is nonlinear, typical methods for joint analysis may become computationally prohibitive due to the intractable high-dimensional integrals in the likelihood. I am developing new models and computational methods to address this computational challenge. Second, the longitudinal data may be used to predict the time to event in a prediction model. This is a new research area called dynamic prediction, as the prediction may change over time as more longitudinal data become available. Various issues need to be studied, including model formulation, estimation and prediction error assessment.
Education & Training
Ph.D. - University of Wisconsin-Madison - 2003