BEGIN:VCALENDAR VERSION:2.0 PRODID:-//wp-events-plugin.com//7.4.0.1//EN TZID:Europe/Paris X-WR-TIMEZONE:Europe/Paris BEGIN:VEVENT UID:142@isdm.umontpellier.fr DTSTART;TZID=Europe/Paris:20250325T160000 DTEND;TZID=Europe/Paris:20250325T160000 DTSTAMP:20260602T130233Z URL:https://isdm.umontpellier.fr/events/combining-t-learning-and-dr-learni ng-a-framework-for-oracle-efficient-estimation-of-causal-contrasts-2/ SUMMARY:Combining T-learning and DR-learning: a framework for oracle-effici ent estimation of causal contrasts DESCRIPTION:Room Nadir\, Maison de la Télédétection (500 rue Jean Franç ois Breton)\nMachine Learning in Montpellier\, Theory &\; Practice - La rs Van der Laan\n\nWe introduce efficient plug-in (EP) learning\, a novel framework for the estimation of heterogeneous causal contrasts\, such as t he conditional average treatment effect and conditional relative risk. The EP-learning framework enjoys the same oracle-efficiency as Neyman-orthogo nal learning strategies\, such as DR-learning and R-learning\, while addre ssing some of their primary drawbacks\, including that (i) their practical applicability can be hindered by loss function non-convexity\; and (ii) t hey may suffer from poor performance and instability due to inverse probab ility weighting and pseudo-outcomes that violate bounds. To avoid these dr awbacks\, EP-learner constructs an efficient plug-in estimator of the popu lation risk function for the causal contrast\, thereby inheriting the stab ility and robustness properties of plug-in estimation strategies like T-le arning. Under reasonable conditions\, EP-learners based on empirical risk minimization are oracle-efficient\, exhibiting asymptotic equivalence to t he minimizer of an oracle-efficient one-step debiased estimator of the pop ulation risk function. In simulation experiments\, we illustrate that EP-l earners of the conditional average treatment effect and conditional relati ve risk outperform state-of-the-art competitors\, including T-learner\, R- learner\, and DR-learner. Open-source implementations of the proposed meth ods are available in our R package hte3.\n\nLabéliséHallesIA ATTACH;FMTTYPE=image/jpeg:https://isdm.umontpellier.fr/wp-content/uploads/ 2025/02/ml-mpt-1.jpg END:VEVENT BEGIN:VTIMEZONE TZID:Europe/Paris X-LIC-LOCATION:Europe/Paris BEGIN:STANDARD DTSTART:20241027T020000 TZOFFSETFROM:+0200 TZOFFSETTO:+0100 TZNAME:CET END:STANDARD END:VTIMEZONE END:VCALENDAR