About

Dr. Paul Zivich is a research assistant professor in the Department of Epidemiology at University of North Carolina at Chapel Hill. His interests are cover quantitative epidemiologic methods development and their practical application to critical public health questions. The main focus of his work is in causal inference with potential outcomes, infectious disease prevention, pharmacoepidemiology, and computational aspects of epidemiology. His work has ranged from assessing the performance of estimators through simulation studies to free and open source software (FOSS) to collection of contact network data with electronic sensors to application of causal inference in the context of infectious disease and social epidemiology.

Paul received his PhD in epidemiology from University of North Carolina at Chapel Hill, and his MPH in epidemiology from The Ohio State University. He was born, is conducting research, and will die.

This website highlights select aspects of my work, software I have written, and other asides. For a full summary of my previous work, please see my CV below.

To adhere to the FAIR (Findability, Accessibility, Interoperability, and Reuse) principles for my research software, code from selected publications is available on Zenodo. For reproducibility, I strive to have aspects of my research work double-coded. Many of my projects have code available in at least two programs (generally, Python with either accompanying SAS or R code). A special thank you to my co-authors who make this possible.

Explainers

A collection of previous explainers I have done

Current Research

The following is a brief summary of my ongoing research projects or current research interests.

Non-standard causal inference

When first taught causal inference, many associate the framework directly with rote memorization of the standard identification assumptions: causal consistency, exchangeability, and positivity. Others associate it with specific estimators (e.g., inverse probability weighting). However, causal inference is so much more. I personally view it as a way to reason about the world that has application to learning from any sort of data (or information).

To showcase this aspect, a major area of my research is in what I will refer to as ‘non-standard’ causal inference. This work is meant to go beyond the basic assumptions and estimators. My work in this area includes synthesis modeling to address violations of the positivity assumption, use of proximal causal inference to account for unmeasured confounding variables, data fusion to integrate information across data sources, machine learning to weaken modeling assumptions, and causal inference with interference.

Estimating equations

During my postdoc, Stephen Cole recommended I learn about M-estimation. Since that time, it has changed how I approach quantitative research. To me, estimating equations present a set of powerful tools for the conduct of epidemiologic research. They have simplified estimation for practical problems, allowed me to design more extensive simulation studies for novel estimators, and easily develop proofs for asymptotic properties of estimators. Most relevant for epidemiologists, they have allowed me to nearly always avoid having to use the bootstrap for variance estimation. My ongoing interest is in this area is furthering the development of these tools and to ease their adoption by epidemiologists.

My most meaningful contribution in this area (in my opinion) has been the development of delicatessen, a Python library for general application of estimating equations. This has enabled and accelerated my other work in this area, which includes introductory materials developed for epidemiologists, extensions for g-computation with marginal structural models or time-varying confounding, use with sensitivity analyses, fusion designs, and bridged treatment comparisons for comparing across randomized trials.

I also conduct workshops on the use of estimating equations. Materials for this workshop are publicly available here.

Infectious diseases

My applied work focuses mainly on infectious disease epidemiology. Previous and ongoing research have focused on HIV, influenza, pneumonia, SARS-CoV-2, testing strategies, and vaccines.

Paul Zivich