Details

Titel: The role of instrumental variables in causal inference

Vortragende: Nataliya Sokolovska (Laboratory of Computational and Quantitative Biology, Sorbonne University, Paris, France)

Wann? Mittwoch, 16. April, 14:00 - 15:00 Uhr
Wo? HS 2, Währinger Straße 29, 1090 Wien

Abstract:
One of the most topical tasks in the modern empirical sciences and data-driven domains is causal inference purely from observational data, in particular, in applications where interventions are not possible. An interesting question is whether causal inference can be done in a bivariate case (for two variables only). Causal inference methods based on conditional independence construct Markov equivalent graphs, and cannot be applied to bivariate cases. The approaches based on independence of cause and mechanism state, on the contrary, that causal discovery can be inferred for two observations. We challenge to reconcile these two research directions. We study the role of latent variables such as latent instrumental variables and hidden common causes in the causal graphical structures. We show that the methods based on the independence of cause and mechanism indirectly contain traces of the existence of the hidden instrumental variables. I will present our novel algorithm to infer causal relationships between two variables, which was validated on simulated data and on a benchmark of cause-effect pairs.

Bio:
Nataliya Sokolovska  is a professor (since 2022) in Computer Science at Sorbonne University. She holds a PhD degree from Telecom ParisTech (2010). She was a post-doctoral researcher at the Computer Science Laboratory at Paris XI University (2010 - 2011) and at the Computing Department, Macquarie University, Sydney, Australia (2011 - 2012). She was assistant professor at Sorbonne University affiliated with the NutriOmics team, Faculty of Medicine (2012 – 2022). Her research interests include graphical models, probabilistic inference, causal inference, interpretable models, semi-supervised learning, reinforcement learning, applications in medicine and biology.