Nima Hejazi

My research explores how advances in causal inference, statistical machine learning, and computational statistics can empower discovery in the biomedical and health sciences. I focus primarily on the development of model-agnostic, assumption-lean statistical inference procedures, doing so while emphasizing a science-first, translational philosophy that stresses the rich interplay between the applied sciences and statistical methodology: how emerging questions in the former spur advances in the latter, which, in turn, help to refine scientific discoveries. To accomplish this, my work leverages causal inference as a framework to translate scientific questions into precise, causally interpretable statistical estimands, and then aims to obtain inference about these from data by formulating analytic methods that incorporate flexible, adaptive modeling strategies (i.e., machine learning), to avoid imposing restrictions that may not be justified by domain knowledge, and semi-parametric efficiency theory for best-in-class uncertainty quantification. I am also interested in statistical instrumentation---that is, high-performance computing and open-source software and programming---to push the boundaries of statistical methodology and to promote transparency and reproducibility in the practice of applied statistics and data science.

My methodological work draws upon tools and ideas from semi- and non-parametric statistics, high-dimensional and large-scale inference, de-biased or targeted machine learning (e.g., targeted minimum loss estimation, sieve estimation), and computational statistics. Areas of recent focus include the study of inference on treatment effects from data collected via biased or outcome-dependent sampling designs, including extensions to sequentially adaptive sampling schemes; causal effect heterogeneity for optimal treatment regime and subgroup discovery; efficient semi-parametric or causal machine learning approaches for evaluating dose-response phenomena; causal mediation analysis (i.e., path-specific direct and indirect effects) for investigating questions of mechanism; and safely drawing causal inferences from data exhibiting network dependence or interference structures.

Inspired by John Tukey's sentiment that "the best thing about being a statistician is that you get to play in everyone's backyard", my past substantive collaborations have spanned diverse areas of the biomedical and public health sciences---from toxicology and computational biology to environmental health and nutritional epidemiology. Recently, though, I've been captivated by the rich scientific and statistical problems that abound in the infectious disease sciences, especially in efforts to evaluate investigational therapeutics and preventive vaccines in clinical trials and observational studies. My work has contributed novel methods and insights for characterizing immune correlates (surrogate endpoints) in vaccine efficacy trials of HIV and COVID-19; for comparing therapeutics in studies of COVID-19 and TB/HIV co-infection; and for identifying post-acute sequelae of COVID-19.

To close, here are a few choice reflections on the intertwined philosophies of science and of statistics:

"Far better an approximate answer to the right question, which is often vague, than the exact answer to the wrong question, which can always be made precise." --John Tukey

"Everyone is sure of this [that errors are normally distributed]...since the experimentalists believe that it is a mathematical theorem, and the mathematicians that it is an experimentally determined fact." --Henri Poincare

"The anarchy of guess and intuition has given way to a benevolent tyranny of statisticians." --Donald Fredrickson

"Science is the belief in the ignorance of experts." --Richard Feynman