Short course: Causal Time-to-Event Analysis
Instructors: Stijn Vansteelandt, Ghent University; Oliver Dukes, Ghent University, University of Pennsylvania; Torben Martinussen, University of Copenhagen
Sponsor: ASA Lifetime Data Science Section
Dates and Times (This course will be taught via Zoom on 2 consecutive days):
Part 1: November 29, 2021: 12-2:30pm ET
Part 2: November 30, 2021: 12-2:30pm ET
Registration Deadline: Wednesday, November 24, at 12:00 p.m. Eastern time
Course Description: Evaluating treatment effects using observational data increasingly requires adjustment for high-dimensional set of covariates in order to control confounding. This is the result of a lack of comparability between treated and untreated subjects in possibly many (pre-treatment) factors that are also related to outcome. Such adjustment is routinely achieved via parametric modelling in a manner that is often not directly targeting the causal question of interest. This can lead to bias (e.g., due to model misspecification or poor detection of confounders), inefficiency and invalid inference (e.g., as a result of ignoring the uncertainty induced by variable selection). In this course, we will review targeted strategies for inferring the causal effect of an exposure on a time-to-event endpoint subject to censoring. On the first day, we will review efficient analysis strategies for randomised experiments, as well as robust methods for analysing observational data. Special attention will be given to the problem of selection and modelling of variables that are sufficient to adjust for confounding and to render censoring non-informative; we will in particular introduce double/triple variable selection strategies, as well as targeted learning techniques. On the second day, we will introduce instrumental variable methods for time-to-event data, as well as methods for inferring causal mechanism based on the mediational g-formula as well as natural effect models. Software demonstrations in R will be used throughout the course.
The course is aimed to be accessible to applied statisticians, epidemiologists, and other quantitative researchers already familiar with standard survival analysis methods as well as with the language of potential/counterfactual outcomes and with identification (i.e., exchangeability or ignorability assumptions), for example for the Average treatment effect.
Registration:
https://www.amstat.org/EventDetail?Eventkey=WS202118
Fees:
LiDS Section Members: $40
Regular ASA Members: $55
Students: $40
Nonmembers $80
Access Information
Registered persons will be sent an email the afternoon of Wednesday, November 24, with the information to join the webinar and, if possible, a link to download and print a copy of the presentation slides.