Fachvortrag in englischer Sprache

Causality and statistics have been discussed in statistical applications for a century, and solutions have been found by the invention of structural equation models, which generate graphical models. We consider a new structural equation model, where all random variables can be written as a max-linear function of their parents and independent noise terms such that the stochastic dependence structure of the corresponding vector can be modeled by a directed acyclic graph. We investigate the properties of our new causal model, for instance, minimal representations of our model. We also discuss various applications and statistical estimation problems.

Speaker: Prof. Dr. Claudia Klüppelberg, TU München