Program
This course is a whistle-stop tour of the key concepts and methods in causal inference, suitable for newcomers to the area. The emphasis throughout is more strongly on the concepts than the methods, with the aim of giving participants enough familiarity with the topics that they are able to follow the majority of the talks in the main EuroCIM meeting. The course will touch on the following:
Causal language (do-notation, potential outcomes, counterfactuals)
Identification, and assumptions that make identification possible (conditional exchangeability / no unmeasured confounding, consistency, positivity, no interference)
Non-parametric and parametric estimation (including the role of traditional regression models in causal inference)
Graphical methods (DAGs and SWIGs)
The propensity score and associated methods
Instrumental variables (including Mendelian randomisation and the regression discontinuity design)
Longitudinal causal effects and time-dependent confounding
Mediation analysis
The course will be delivered via a combination of pre-recorded short videos, “live” video lectures, a small-group discussion and a short practical session (which can be completed using any statistical software). Although the pre-recorded videos will be made available in advance, there will be breaks during the course for these to be watched.
To allow interactions and group discussions the number of participants is limited to 110 on a first come first served basis.
Timetable (UK time)
9-9:15 INTRODUCTION (“Live”)
9:15-10 PRE-RECORDED VIDEOS (Causal language, identification, non-parametric estimation)
10-10:45 LIVE LECTURE (DAGs and SWIGs)
10:45-11 BREAK
11-11.45 LIVE LECTURE (Parametric estimation using regression models, the propensity score)
11.45-12.30 SMALL GROUP DISCUSSION
12:30-1:30 LUNCH BREAK
1.30-2 LIVE LECTURE (instrumental variables)
2-2:45 PRE-RECORDED VIDEOS (longitudinal causal effects, mediation analysis)
2:45-3 BREAK
3-4 PRACTICAL SESSION
