Dr. Hong earned a Ph.D. in statistics from the University of Illinois at Urbana-Champaign. As a statistical scientist, she focuses on the development of cutting-edge statistical methods for analyses of complex large-scale datasets (e.g., high-dimensional censored data and longitudinal data), and applying these methods to the fields of public health, medicine, and health policy research. Prior to joining DCEG, Dr. Hong was an associate professor in the Department of Probability and Statistics at Michigan State University in East Lansing. She was awarded scientific tenure by the NIH and joined the Biostatistics Branch as a senior investigator in August 2021.
In her role in DCEG, Dr. Hong works to advance scientific knowledge in population-based cancer epidemiology and genetics studies through the development and use of novel statistical methodology. She has led the development of methods in high-dimensional time-to-event analysis by proposing a series of novel and innovative ideas. She has made important advances in statistical theory and methodological development in the areas of quantile regression analysis, classification with high-dimensional features, and high-dimensional time-to-event analysis. Quantile regression has emerged as both an efficient way of linking the whole distribution of an outcome to the covariates of interest and an important alternative to commonly used regression models. Dr. Hong has led pioneering work to address the many challenges inherent to this approach, in terms of theory and computation when the covariates of interest are high dimensional.
Dr. Hong has also contributed to statistical methodology for classification problems with high-dimensional covariates. For example, many classification problems emerge from analyses of gene expression and imaging data to identify individuals with disease or at high risk of developing disease. To circumvent these issues, she proposed a novel high-dimensional classification method that could predict clinical diagnosis of autism spectrum disorder using imaging predictors, integrating priorly known different sources of anatomical information, correlation among imaging predictors, and spatial information.