Compliance and causal analysis

Aim

To present recent developments in biostatistics of methods for extending conventional modelling of longitudinal studies to more properly address issues of confounding, selective exposure, and intermediate response dependent behaviour. A main initiator in this development is James M. Robins, Harvard University, but important contributions have also come from econometrics. Closely related techniques have been developed for extending the conventional intention-to-treat analysis of clinical trials to 'as treated' under less than full compliance in such a way that the direct 'causal' effect is identified. The purpose of this course is to introduce some of these ideas and techniques (basic counterfactual ideas, propensity scoring and inverse probability of treatment weighting, double robust estimators, instrumental variables, structural means models, marginal structural models and G-estimation). A main focus will be computer classes using the implementation of the methods in STATA, using a variety of procedures including gllamm.

Location & Organization

Organizer

DGSB - Danish Graduate School in Biostatistics

Course Director

Niels Keiding

Location / venue

University of Copenhagen (Faculty of Health Sciences)

Timing & Workload

Duration 3 days
ECTS points 3
Frequency Every third year
 

Examination yes

Criteria

Is the course taught in English? yes
Is documentation available? (book, syllabus)? yes
Is the course open for external researchers? yes

More Information

http://pubhealth.ku.dk/bs/undervisning/phdkurser/efteraar2006/compliance/