The focus of this research unit is on applying state-of-the-art statistical and econometric methods to applied questions with focus on health economics, insurance, public health and epidemiology. In most of our work, we combine empirical questions with theoretical contributions in terms of novel identification strategies or advances in econometric modelling.
Technological developments in the last years, like the evolution of the internet and the increasing digitalization of many areas of life, have caused, that data concerning human behavior and the interaction of human beings are available in size and quality which have never been seen before and are in principle available for scientific research. While on the one hand these data sets enable new insights and allow addressing new research questions, they on the other hand impose new methodological challenges. Firstly, it is computationally challenging to handle the size of the data. Secondly, "classical" statistical methods are often not suitable to learn from complex big datasets.
Machine learning is an increasingly popular topic. Several disciplines can benefit from machine learning techniques. While the theoretical properties of machine learning techniques are now very well investigated in econometrics and statistics, these methods are still not very often used in practice. In this research unit we will mainly focus on a particular machine learning technique: the Least Absolute Selection and Shrinkage Estimator (Lasso). This technique is particularly promising in economic applications as its selection part simplifies the interpretation of complex economic models.