Fan Li, PhD
Associate Professor of Biostatistics and Associate Professor of Medicine (Cardiovascular Medicine)Cards
About
Research
Overview
Below is an overview of my research agenda and contributions. I welcome opportunities to collaborate with researchers from diverse disciplines and invite strong students and postdoctoral scholars to reach out regarding potential research opportunities. I am particularly interested in working with students who have strong mathematical foundations and/or exceptional computing skills. I have had the privilege of collaborating with many graduate students, postdoctoral researchers, and faculty colleagues, whose insights and dedication have played a crucial role in shaping and advancing our research projects together.
Cluster Randomized Trials & Stepped Wedge Designs: My team develops statistical methods to enhance the design and efficiency of cluster randomized trials (CRTs), including parallel-arm, crossover, stepped wedge, and individually randomized group treatment designs. We are interested in expanding the statistical toolbox for CRTs to handle missing covariate and outcome data, to properly address complex survival and composite endpoints, and to enable detection of measured and unmeasured treatment effect heterogeneity. We are also advancing an estimand-aligned and efficiency-focused framework, contributing both theoretical advancements and applied innovations that shape modern CRT and pragmatic clinical trial research.
Propensity Score/Principal Score Methods: My team develops statistical methods to improve causal effect estimation in both randomized and observational studies. Our work focuses on advancing propensity score weighting—for example, the overlap weighting approach, and other semiparametric efficient techniques to address confounding and selection bias. We are also extending principal stratification methods—for example, the principal score approach, to handle intermediate variables (also referred to as "intercurrent events" under the ICH E9 Estimands Framework), addressing challenging such as censored survival outcomes, multiple treatments, clustering, and non-monotonicity.
Causal Mediation Methods to Uncover Mechanisms: My team develops statistical methods for regression and causal mediation analysis in complex biomedical settings, including mismeasured exposures, survival endpoints, complex mediators, and post-treatment confounding. We focus on developing robust and efficient estimators and sensitivity analysis techniques that strengthen inference under transparent and interpretable assumptions. Our recent interest includes advancing semiparametric efficient and flexible outcome modeling methods for studying natural mediation effects and spillover mediation effects within a multilevel data context, to improve our understanding of causal mechanisms in modern pragmatic and implementation trials.
De-biased Causal Machine Learning & Bayesian Causal Machine Learning: My team studies principled methodologies that integrate modern causal machine learning to improve robustness, efficiency, and interpretability in clinical trials and observational studies. Two typical examples include (a) de-biased machine learning methods that integrate flexible nuisance function estimators for estimating the average causal effects, quantile causal effects, natural direct and indirect effects and principal causal effects; (b) Bayesian additive regression trees methods that flexibly model complex outcome surfaces for addressing conditional average causal effects, conditional principal average causal effects, and time-varying causal effects.
Statistical Software Development: My team has been developing open-source statistical software to support causal inference, (clustered or non-clustered) clinical trials, observational studies, and mediation analysis. Examples of our past contributions include R packages for propensity score methods (e.g., PSweight), variance estimation with finite-sample adjustments (e.g., CoxBCV), mediation analysis (e.g., mediateP), constrained randomization (e.g., cvcrand), and marginal modeling (e.g., geeCRT), among others. We focus on disseminating computational tools that facilitate the practical implementation of new statistical methods, making a small but meaningful effort to translate methodological innovations into practice change for clinical, biomedical, and public health research.
Empirical Research & Interdisciplinary Collaborations: While my team consists primarily of biostatisticians, we have a strong interest in collaborating with researchers from diverse backgrounds. We have worked on multiple systematic reviews on pragmatic clinical trials, fostering a deep appreciation for empirical research as a means of understanding the current landscape of statistical and methodological practice. These efforts have largely motivated my work in addressing complex statistical challenges that are factually grounded and well-motivated. My team actively collaborates with clinical scientists and epidemiologists across a wide range of disciplines, including but not limited to cardiology, nephrology, aging and dementia, emergency medicine, critical and intensive care, and implementation science.
Medical Research Interests
Public Health Interests
Academic Achievements & Community Involvement
News & Links
News
- December 11, 2024
Paradigm-shifting work brought biostatistics student to YSPH
- October 25, 2024
Can a ‘Kidney Action Team’ Improve Patient Outcomes?
- October 24, 2024
New Analytics Center for Cardiovascular Medicine
- February 05, 2024
Patient Priorities Care Shows Potential for Improving Outcomes for Older Adults With Multiple Chronic Conditions