Statistics/Finance Seminar with Christine Bang, University of Copenhagen

The Department of Finance and Center for Statistics are proud to announce the upcoming seminar with Christine Bang, University of Copenhagen.

Christine Bang will present: Causal discovery with temporal background knowledge

Abstract:
Constraint-based causal discovery methods aim to recover causal graphs from conditional independencies among the variables. Multiple causal structures may induce the same dependence structure and form an equivalence class. Without additional, stronger assumptions, it is usually not possible to recover more than the equivalence class, i.e. we cannot identify all causal directions. Moreover, constraint-based methods are sensitive to errors based on statistical testing when applied to real data. In practice, we often have more causal information available than what can be obtained from the data alone. Time structured data induces a partial causal ordering of the variables, which I will refer to as temporal, or tiered, background knowledge. This type of background knowledge is common, e.g. for cohort data, and it improves causal discovery methods such that the output becomes more informative and reliable. In this talk, I will first give a brief introduction to constraint-based causal discovery algorithms and show how temporal background knowledge can be incorporated. Then, I will show how this improves graphs estimated using finite sample data. In addition, I will show which time orderings yield the most informative graphs and that the output graphs have some desirable theoretical properties. Finally, I will show that temporal background knowledge can also be used to improve causal discovery if we allow for latent confounding, and if we consider multiple datasets with overlapping variables.

Location:
Solbjerg Plads 3
2000 Frederiksberg
Room: SPs03