Séminaire de Santé publique
Jeudi 11 juin 2026 - 13h à 14h
Amphi Louis - ISPED
Campus Carreire - université de Bordeaux
Ouvert à tous
En présentiel et visioconférence
Seminar in English
Title: Missing data in epidemiological studies. What to do when the usual assumptions fail.
Constantin Theodore Yiannoutsos, Ph.D.
Professor of Biostatistics
City University of New York
Graduate School of Public Health and Health Policy
Bio here
Abstract: Missing data in epidemiological studies seriously complicate the analysis and render conclusions potentially invalid. All methods proposed for handling missing data are based on the assumption that what you see is the same as what you don’t see (data missing completely at random – MCAR – or at random – MAR). When data are missing not at random (MNAR), what you don’t see cannot be extrapolated from available data. What’s more, available data are insufficient to guide you as to which type of missingness you are dealing with. I will present a number of methods, based on strategically augmented observational data, which attempt to address the MNAR issue by recovering MAR in the combined data. I will apply these methods in analyses of time to event (also known as “survival” analyses), models of competing risks and in longitudinal data. I will also present some recent developments about how to perform principled sensitivity analyses when all else fails.
