Associate Adjunct Professor, UCSF School of Dentistry
Dr. Cheng earned her medical degree in China, M.S. in nutrition at Cornell University in 2002, and Ph.D. in biostatistics at the University of Pennsylvania in 2006. Before joining UCSF School of Dentistry in 2010, Dr. Cheng was an Assistant Professor in Biostatistics at the University of Florida’s College of Medicine.
Dr. Cheng develops new statistical methods for complex problems in randomized trials and observational studies and is experienced in working with investigators in various fields of health sciences including dentistry and oral diseases, biomedicine, infectious diseases, pharmacogenomics, nursing, and public policy research. Dr. Cheng is currently a principal investigator (PI) on mediation analysis for dental studies at CANDO, a project to develop statistical methods to better understand how mechanisms of an intervention function and which components are more likely to succeed or fail. This research can help to understand the causal pathway/mechanism of an intervention and improve future programs by tailoring specific components of an intervention.
Dr. Cheng is also a statistician at CANDO’s Data Coordinating Center (DCC) and UCSF CTSI, where she works with investigators on study design, power analysis, randomization, statistical analysis, and the preparation of grant proposals and manuscripts.
Dr. Cheng's research interests include causal inference (instrument variables and propensity scores) with applications in clinical trials with complex issues, e.g., noncompliance, mediation with intermediate variables (biomarkers, attitude, knowledge, behaviors etc.), missing data, and outcome only observed in “survivors” etc., and in observational studies with measured and unmeasured confounding, methods for genetic association studies, categorical data analysis, longitudinal data analysis, survey design and analysis, and nonparametric statistics.
Accurate assessment of the impact of tobacco use on health outcomes and healthcare expenditures is critical in regulatory science to assist policy makers to promulgate rational policies. However, confounding is a potential problem for assessments based on observational studies. Recently, modern analytic methods that use causal modeling have been successfully applied to address confounding inherent in observational research.