CAST Seminar: Janie Coulombe

Join us at the CANSSI Ontario STatistics Seminars (CAST) with

Janie Coulombe

Assistant Professor
Department of Mathematics and Statistics
Université de Montréal

Free Event | Registration Required

Talk Title

A novel multiply robust estimator for the average treatment effect that addresses confounding and covariate-driven observations

Abstract

Randomized controlled trials (RCTs) are the gold standard for drawing causal inferences about the marginal effect of a treatment. When it is not possible to conduct an RCT, we often turn to observational data, which are not meant for research purposes and present with different features that can affect the causal inference when they are not considered adequately. In this talk, I focus on two such features or challenges which are the confounding and covariate-driven monitoring times. These features can affect the inference by inherently creating spurious associations between the treatment and the outcome of interest. I describe that issue in the context of a longitudinal study using data from electronic health records. Using semiparametric theory, I then propose a novel, efficient estimator for the causal effect of treatment that accounts for informative monitoring times and confounding. In addition to being less variable than a previously proposed alternative estimator, the novel estimator is multiply robust to misspecification of different nuisance models used in its construction. It is demonstrated theoretically and in extensive simulation studies. We then apply it to data from the Add Health study in the US to study the causal effect of psychotherapy on alcohol consumption. This is joint work with Professor Shu Yang at North Carolina State University.

Speaker Profile

Dr. Janie Coulombe is an Assistant Professor of Statistics in the Department of Mathematics and Statistics at Université de Montréal. Her research specializes in the development of causal estimators with robust statistical properties to address challenges inherent in observational data, such as irregular measurement times and confounding. In addition to her methodological contributions, Dr. Coulombe has built extensive experience as a data analyst at the Lady Davis Research Institute in Montreal, where she collaborated with experts in pharmacoepidemiology between 2014-2019. Her work there involved analyzing large electronic health records and administrative databases, an experience that has significantly influenced her current methodological research program at Université de Montréal.