Causal Inference: Johannes Buck
From Belle Taylor on July 2nd, 2021
This talk has been automatically captioned. You can remove these by pressing CC on the video toolbar.
Name: Johannes Buck
Talk Title: Properties and Consistency of QTree in Max-Linear Models Under Observational Noise
Abstract: Recently, we proposed the QTree algorithm for causal inference in river networks. Motivated by max-linear models, the algorithmn utilizes the quantile gap between differences of random variables to unravel causal relationships and achieves almost perfect recovery of the upper Danube basin, clearly outperforming existing methods.
In this talk, we present the qualitative features of the estimator. Assuming that the data comes from a max-linear Bayesian network with additive noise, we prove that QTree is consistent given some mild conditions on the tail of the noise distribution. This is joint work with Ngoc Mai Tran and Claudia Klüppelberg.
This talk is a contributed talk at EVA 2021.