Benedicte Colnet, a Ph.D. student under the supervision of Julie Josse, Gaël Varoquaux, and Erwan Scornet at Inria (France), will present her work on "Reweighting the RCT for generalization: finite sample error and variable selection".
Overview
The event has passed
- Date:Starts 17 April 2023, 14:00Ends 17 April 2023, 15:00
- Location:Analysen, EDIT building
- Language:English
Abstract
Randomized Controlled Trials (RCTs) may suffer from limited scope. In particular, samples may be unrepresentative: some RCTs over- or under-sample individuals with certain characteristics compared to the target population, for which one wants conclusions on treatment effectiveness. Re-weighting trial individuals to match the target population can improve the treatment effect estimation. This talk will present the questions at hand with a clinical example from critical care. Then, we present recent guarantees on one of the main estimators: the Inverse Propensity of Sampling Weighting (IPSW). First, we present the similarities and differences with the typical Inverse Propensity Weighting (IPW) estimator in causal inference. Second, we propose a finite and large sample characterization of IPSW when the covariates are categorical. Our results show how the performance (bias, variance, and quadratic risk) of IPSW estimates depends on two sample sizes (RCT and target population sample to build the weights). Results also reveal that IPSW performances are improved when the trial probability to be treated is estimated (rather than using its oracle counterpart). In addition, we study choice of variables: how including covariates that are not necessary for identifiability of the causal effect may impact the asymptotic variance. Including covariates that are shifted between the two samples but not treatment effect modifiers increases the variance while non-shifted but treatment effect modifiers do not.
About the speaker
Benedicte Colnet is an engineer by training and is a third-year PhD student under the supervision of Julie Josse, Gaël Varoquaux, and Erwan Scornet at Inria (France). She is working on causal inference, with a focus on randomized controlled trial's representativeness. As she works with clinicians, her work usually addresses clinical questions.
DSAI Seminars
This is a seminar from the DSAI seminars series held every Monday at 14:00 by the Data Science and AI division. The seminars are usually hybrid.
