Prof. Dr. Andreas Mayr

Our research aims at developing new statistical methods and software for medical data analysis. The focus is on advanced statistical regression models: Particularly in epidemiological research, where one typically deals with observational data without randomization, one needs to adjust estimators for possible confounding variables via statistical models. Due to growing amounts of data, classical selection of risk factors or confounders by subject knowledge only is often no longer possible. Therefore, we develop statistical algorithms that are able to automatically select the most influential variables out of a potentially large data set based on statistical learning approaches. Additionally, we focus on model classes, that go beyond the classical regression of the mean by focusing on the complete conditional distribution and are more flexible with respect to the shape of effects.