BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Causal discovery through Pearson risk invariance - Veronica Vincio
 tti (University of Trento)
DTSTART:20260305T103000Z
DTEND:20260305T111500Z
UID:TALK244411@talks.cam.ac.uk
DESCRIPTION:Prediction invariance of causal models under heterogeneous set
 tings has been exploited by a number of recent methods for causal discover
 y\, typically focussing on recovering the causal parents of a target varia
 ble of interest. Existing methods require observational data from a number
  of sufficiently different environments\, which is rarely available. In th
 is work\, we consider a structural equation model where the target variabl
 e is described by a generalized linear model conditional on its parents. U
 nder this setting\, we characterize the causal model via Pearson risk inva
 riance\, leading to a computational strategy for searching the causal mode
 l among all possible models. Crucially\, for generalized linear models wit
 h a known dispersion parameter\, such as Poisson and logistic regression\,
  the causal model can be identified from a single data environment. Applic
 ations range from classical Poisson/logistic regression association studie
 s\, now seen through a causal lens\, to relational event models for dynami
 c network data. The method is implemented in the R package causalreg.
LOCATION:Seminar Room 1\, Newton Institute
END:VEVENT
END:VCALENDAR