BEGIN:VCALENDAR VERSION:2.0 PRODID:-//132.216.98.100//NONSGML kigkonsult.se iCalcreator 2.20.4// BEGIN:VEVENT UID:20260119T005446EST-4142NLEGIK@132.216.98.100 DTSTAMP:20260119T055446Z DESCRIPTION:JOINT CORE/EBOH EPI Seminar Series Winter 2026\n\nThe Seminars in Epidemiology organized by the Department of Epidemiology\, Biostatistic s and Occupational Health at the ɬÀï·¬ School of Population and Global He alth is a self-approved Group Learning Activity (Section 1) as defined by the maintenance of certification program of the Royal College of Physician s and Surgeons of Canada. Physicians requiring accreditation\, please comp lete the Evaluation Form and send to admincoord.eboh [at] mcgill.ca.\n\nPh ilippe Boileau\, PhD\n\nAssistant Professor of Biostatistics\n Department o f Epidemiology\, Biostatistics and Occupation Health\n ɬÀï·¬\n \nWHEN: Monday\, JANUARY 19\, 2026\, from 3:30-4:30pm\n WHERE: Hybrid | Ons ite at 5252 boul. de Maisonneuve - 3rd floor\, 3B Kitchen | Zoom\n Note: Ph ilippe Boileau will be presenting in-person at CORE\n\nAbstract\n\nThe con ditional average treatment effect (CATE) is frequently estimated to refute the homogeneous treatment effect assumption. Under this assumption\, all units making up the population under study experience identical benefit fr om a given treatment. Uncovering heterogeneous treatment effects through i nference about the CATE\, however\, requires that covariates truly modifyi ng the treatment effect be reliably collected at baseline. CATE-based tech niques will necessarily fail to detect violations when effect modifiers ar e omitted from the data due to\, for example\, resource constraints. Sever e measurement error has a similar impact. To address these limitations\, w e prove that the homogeneous treatment effect assumption can be gauged thr ough inference about contrasts of the potential outcomes’ variances. We de rive causal machine learning estimators of these contrasts and study their asymptotic properties. We establish that these estimators are doubly robu st and asymptotically linear under mild conditions\, permitting formal hyp othesis testing about the homogeneous treatment effect assumption even whe n effect modifiers are missing or mismeasured. Numerical experiments demon strate that these estimators’ asymptotic guarantees are approximately achi eved in experimental and observational data alike. These inference procedu res are then used to detect heterogeneous treatment effects in the re-anal ysis of randomized controlled trials investigating targeted temperature ma nagement in cardiac arrest patients.\n\nSpeaker Bio\n\nPhilippe Boileau is an Assistant Professor of Biostatistics at ɬÀï·¬ with a joint appointment in the Department of Epidemiology\, Biostatistics and Occupat ional Health and the Department of Medicine. He is also a Junior Scientist at the Research Institute of the ɬÀï·¬ Health Centre\, where he is the director of the Novel Trial Methods Hub. Dr. Boileau is broadly interested in the development of assumption-lean statistical methods and t heir application to quantitative problems in the health and life sciences. Assumption-lean procedures combine causal inference\, machine learning\, and semiparametric techniques to provide dependable statistical inference without relying on convenience assumptions. His most recent work has focus ed on developing and applying causal machine learning methods for heteroge neous treatment effect discovery in clinical trial data.\n\nLearning Objec tives\n\nAt the completion of this talk\, attendees will be able to:\n\n\n Describe limitations of existing statistical methods for heterogeneous tre atment effect detection\;\n Interpret differential variance parameters\;\n T ranslate treatment effect homogeneity test results to clinical contexts.\n \n DTSTART:20260119T203000Z DTEND:20260119T213000Z SUMMARY:Assumption-Lean Differential Variance Inference for Heterogeneous T reatment Effect Detection URL:/spgh/channels/event/assumption-lean-differential- variance-inference-heterogeneous-treatment-effect-detection-369792 END:VEVENT END:VCALENDAR