Hosted by the GPH Department of Biostatistics and the NYU Grossman School of Medicine’s (GSOM) Division of Biostatistics
In this seminar, GSOM’s Dr. Mengling Liu will present a family of dynamic single-index models to address the gap between existing analytical methods and a recent increased interest in the study of time-independent multiple exposure mixtures. Additionally, she will introduce a nonparametric hypothesis testing procedure to assess whether the effects of the exposure mixture vary over time, and a model averaging approach to mitigate issues related to knot selection in spline approximations.
Abstract: Environmental exposures often display temporal variability, driving significant research to understand their dynamic effects on human health. Recently, there has been increased interest in studying time-dependent multiple exposure mixtures, extending beyond single-exposure assessments cross-sectionally to account for multi-dimensional exposures over time. However, existing analytical methods typically evaluate exposures individually or assume additive relationships. Dr. Mengling Liu presents a family of dynamic single-index models to address this gap, enabling the evaluation of joint effects of multiple time-dependent exposures on scalar outcomes and time-to-event outcomes. These models capture the time-varying effects of exposure mixtures using a nonparametric bivariate exposure-time-outcome intensity function. Spline tensor product basis functions are employed for nonparametric approximation, and a profile algorithm is proposed for model estimation. Additionally, she introduces a nonparametric hypothesis testing procedure to assess whether the effects of the exposure mixture vary over time. To mitigate issues related to knot selection in spline approximations, she proposes a model averaging approach. The performance of the proposed methods is rigorously evaluated through extensive simulations and demonstrated in multiple applied settings.
About the Speaker
Dr. Liu is a Professor in the Division of Biostatistics at the Department of Population Health of NYU’s Grossman School of Medicine. Her research is primarily focused on developing and applying statistical methods and algorithms for analyzing biomedical data, with the goal of identifying important signals and messages from data to improve human health.
She has led multiple NIH-funded projects as the PI, MPI, or sub-contract PI on developing innovative statistical methods for the analysis of complex survival and longitudinal data, investigating population heterogeneity in pooling projects, and building breast cancer risk prediction models. Currently, she is leading an R01 project on developing statistical methodology for evaluating time-dependent environmental mixtures and a U01 project with Dr. Reibman (MPI) studying the complex WTC exposures' impact on persistent airflow limitation in the WTC Survivor population.
She has a broad range of collaborations with clinical and basic science investigators from multiple departments at NYU Langone, supporting statistical needs in biomedical projects. As the Graduate Advisor for the Biostatistics PhD Program in the Vilcek Institute of Graduate Biomedical Sciences, she work closely with their graduate students on training and research during their entire course in the program.