Causal Reasoning for Data Science

Abstract

In this talk I will illustrate causal reasoning and methodology with two applications involving large databases: First we consider the question of predicting intervention effects in genetic regulatory systems. When the aim is to rank genetic intervention targets according to their expected effects, it can be shown that algorithms taking the causal structure into account outperform standard prediction algorithms. Secondly, we consider the task of assessing the effect of cancer screening programmes from large health claims databases. Here, I will explain how we can avoid self-inflicted biases that arise due to the time-dependent and non-experimental nature of the database by emulating a hypothetical trial.


About the Speaker

In 1996, Vanessa Didelez graduated in the subject of Statistics with Psychology at the University of Dortmund, Germany. After four years as researcher in the special research unit “Discrete Structures”, University of Munich, she received her PhD (Dr. rer. nat.) in Statistics from the University of Dortmund. During 2001-2007 she was a lecturer at the Department of Statistical Science, University College London. Subsequently she moved to the University of Bristol, first as senior lecturer at the School of Mathematics, then promoted to Reader in Statistics; moreover she obtained a Leverhulme Research Fellowship on “Statistical models and methods for complex causal inference”.

In 2016 she was appointed Professor of Statistics with Focus on Causal Inference at the Department of Mathematics and Computer Science at the University of Bremen, as well as being Deputy Head of the Department of Biometry and Data Management at the Leibniz-Institute of Prevention Research and Epidemiology – BIPS in Bremen. Since 2018 she leads the DFG project “Causal Discovery for Cohort Data”.

Speaker:
Prof. Dr. Vanessa Didelez