Olivier Dukes (Université de Gent)
Bespoke instrumental variables for causal inference
Many proposals for the identification of causal effects in the presence of unmeasured confounding require an instrumental variable or negative control that satisfies strong, untestable exclusion restrictions. In this talk, I will instead show how one can identify causal effects for a point exposure by using a measured confounder as a ‘bespoke instrumental variable’. This strategy requires an external reference population that does not have access to the exposure, and a stability condition on the confounder-outcome association between reference and target populations. This is joint work with David Richardson (UC Irvine), Zach Shahn (CUNY) and Eric Tchetgen Tchetgen (UPenn).
L’exposé sera diffusé en ligne au bout de ce lien zoom.