Julemøde i Dansk Selskab for Teoretisk Statistik

Jonas Peters blev tidligere på året tildelt Royal Statistics Society's Guy Bronze medalje for sin forskning. Det ønsker vi at hædre i DSTS-regi.

Titel: Towards using causality in the sciences.

Abstract: In many real world problems, we are interested in knowing how a system responds to interventions. Such interventional effects can be predicted whenever a causal model (rather than a classical statistical model) of the underlying data generating process is available. But even if the research question at hand is not an interventional question, causal ideas may prove to be helpful. When generalizing predictive models to unseen experimental conditions, for example, causal concepts may indicate which part of the model can be assumed to be invariant. Applying causal methods and ideas in the sciences, however, comes with several difficulties: causal knowledge is usually sparse, assumptions are hard to test and the data have complex dependencies. In this talk, we present two simple applications of causal concepts in the sciences: computing causal effects in spatio-temporal data from biogeography and generalizing models of metabolic networks to unseen experiments. No prior knowledge on causality is required.