Projects
Submitted: Treatment Effect Estimation in Causal Survival Analysis: Practical Recommendations
arXiv Preprint | HAL | Code
This article explores causal survival analysis, which combines causal inference and survival analysis to assess the effect of a treatment on time-to-event outcomes in the presence of censoring. Specifically, we focus on estimating the ATE for time-to-event data with static treatment assignment, baseline covariates, and right-censoring.
Status: The project is complete and the manuscript has been submitted to Biometrical Journal.
In preparation (submission soon): Causal Effect of Treatment Duration via RMST in Observational Data
This ongoing work investigates the causal effect of treatment duration in real-world observational settings using Restricted Mean Survival Time (RMST) as the primary estimand.
We develop and compare approaches tailored to time-to-event outcomes with time-dependent treatment patterns and censoring, with the goal of providing practical guidance for applied analyses.
Status: manuscript in final preparation, submission planned soon.
In preparation (soon): Formalizing cloning–censoring–weighting (CCW)
In collaboration with Iqraa Meah, François Petit, and Clément Berenfeld, we are working on methodological and formal aspects of the cloning–censoring–weighting framework.
The objective is to clarify assumptions, define estimands and target trials precisely, and strengthen the theoretical foundations for causal inference on treatment strategies and durations in survival settings.
Status: ongoing.
Simulation study (soon): Variable selection in causal survival analysis
This simulation project focuses on variable selection for causal effect estimation with time-to-event outcomes.
We evaluate the impact of selecting covariates for nuisance models (e.g., censoring and treatment models) on bias/variance trade-offs, robustness, and finite-sample performance across a range of realistic censoring and confounding scenarios.
Status: ongoing.