BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:BSU Seminar: “How to obtain valid tests and confidence intervals
  for treatment effects after confounder selection” - Dr Oliver Dukes\, G
 hent University 
DTSTART:20190122T140000Z
DTEND:20190122T150000Z
UID:TALK115993@talks.cam.ac.uk
CONTACT:Alison Quenault
DESCRIPTION:The problem of how to best select variables for confounding ad
 justment forms one of the key challenges in the evaluation of exposure or 
 treatment effects in observational studies. Routine practice is often base
 d on stepwise selection procedures that use hypothesis testing\, change-in
 -estimate assessments or the lasso\, which have all been criticised for no
 t giving sufficient priority to the selection of confounders. This has pro
 mpted vigorous recent activity in developing procedures that prioritise th
 e selection of confounders\, while preventing the selection of so-called i
 nstrumental variables that are associated with exposure but not outcome (a
 fter adjustment for the exposure). A major drawback of all these procedure
 s is that there is no finite sample size at which they are guaranteed to d
 eliver treatment effect estimators and associated confidence intervals wit
 h adequate performance. This is the result of the estimator jumping back a
 nd forth between different selected models\, and standard confidence inter
 vals ignoring the resulting model selection uncertainty.\n\nIn this talk\,
  I will develop insight into this by evaluating the finite-sample distribu
 tion of the exposure effect estimator in linear regression\, under a numbe
 r of the aforementioned confounder selection procedures. I will then propo
 se a simple and generic solution in the context of generalised linear mode
 ls\, which overcomes this concern (under weaker conditions than competing 
 proposals).
LOCATION:Large Seminar Room\, 1st Floor\, Institute of Public Health\, Uni
 versity Forvie Site\, Robinson Way\, Cambridge
END:VEVENT
END:VCALENDAR