Checking the Positivity Assumption in Practice

Arthur Chatton, PhD

Assistant Professor of Biostatistics
School of Public Health - Université de Montréal 

WHEN: Wednesday, February 11, 2026, from 3:30 to 4:30 p.m.
WHERE: Hybrid | 2001 McGill College Avenue, Rm 1140; Zoom
NOTE: Arthur Chatton will be presenting in-person at SPGH 

Abstract

The positivity assumption, meaning all individuals can receive each treatment modality, is fundamental in causal inference. However, its empirical verification is often overlooked, presumably because of its difficulty. After discussing the different types of violations that may occur, this talk will show how to go beyond propensity score to verify this assumption using the Positivity Regression Trees (PoRT) algorithm available in the R library port. Their advantages and disadvantages will be discussed, along with potential solutions in the case of a violation. If time permits, an extension to the longitudinal and mediation context will be briefly discussed.

Speaker Bio

Arthur Chatton is an Assistant Professor of Biostatistics at the University of Montreal School of Public Health. His works are at the crossroad of prediction and causation, with an emphasis on the use of super learning approaches. He aims to (i) evaluate and improve the quality, usability and generalizability of predictive scores for personalizing medical care and (ii) develop and evaluate new causal methods that consider the imperfection of data collected and used for applied research purposes. For more information, please visit: https://arthurchatton.netlify.app/