
- This event has passed.
Atlantic Canada Data Science Tour
March 7, 2025 @ 12:00 pm - 1:00 pm
A Hybrid Seminar Series for University Students in Statistics or Computer Science
“Mediation Analysis of Recurrent Events”
Guest Speaker: Shenita Pramij (Memorial University)
DATE: Friday, March 7, 2025
TIME: 12:00–1:00 p.m., Atlantic time
LOCATION: Online via ZoomThis hybrid series for students in statistics and computer science programs features faculty members and students from universities in Atlantic Canada talking about their work and career paths in statistical sciences. The guest speaker for this online presentation will be Shenita Pramij.
Shenita is a PhD student in Statistics at Memorial University of Newfoundland. Her research focuses on mediation analysis, and she has broad interests in modelling complex processes with applications in healthcare and public policy. She also has extensive experience in the public sector as a compliance researcher. Her work aims to enhance decision-making in public policy and inform targeted interventions.
PRESENTATION ABSTRACT : Inferring the direct effects of exposure in recurrent event processes, while accounting for mediating factors, is crucial, yet conventional approaches face significant limitations in the presence of complex causal relationships. We introduce two methods to address these challenges. We first explore a two-stage sequential G-estimation method to estimate the controlled direct effect of a randomly assigned exposure, while accounting for potential mediators and confounders, using intensity-based models of recurrent event processes. We also introduce a novel one-stage estimation method based on the estimating equations framework, leveraging the sequential G-estimation principle. We demonstrate that both methods yield unbiased controlled direct effect estimates. The one-stage method also enables the analytical derivation of an estimator for the standard error of the direct effect estimator. We illustrate our approach using a hospital readmission dataset of colorectal cancer patients to estimate the controlled direct effect of sex differences on hospital readmission.
We invite you to join us online! (We’ll send you the Zoom link when you register.)