Tree-based, hybrid regression for modeling biomedical data

• Applicant: Dr. Moritz Berger
• Subject Area: Epidemiology and Medical Biometry/Statistics
• Term: since 2020
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 456235587

Statistical learning techniques play an increasingly important role for the evaluation of clinical and epidemiological data. An objective typically is the modeling of the relationship between a large number of explanatory variables and a dependent variable of interest (response variable). This can be investigated by so-called supervised learning methods. Classical supervised learning methods are generalized linear regression models (GLMs) and tree-based methods (trees). The high importance of these methods is, for example, reflected in numerous studies of the University Hospital Bonn. Subject of the planned project are hybrid regression models, which combine classical GLMs and trees. This approach is particularly promising because it leads to sparse and flexible models that (i) can handle high-dimensional predictor spaces, (ii) can capture possible non-linear effects of the explanatory variables on the response variable, and (iii) can detect relevant interactions between the explanatory variables. The central objective of the research project applied for is to refine the class of hybrid, tree-based regression models, that was already proposed in various preliminary articles. On the one hand, this includes the extension of the model class to new areas of application that are associated with different data situations, and on the other hand, the development of new strategies for stabilizing the model fit and for calculating confidence intervals. In the course of the methodological developments it is planned to make the implementations freely available by suitable software (supplementing the existing program packages). Furthermore, the added value of the proposed methods will be investigated by means of extensive simulations and by using selected clinical and epidemiological data (e.g. from the University Hospital Bonn).

• DFG Programme: Research Grants
• Co-Investigator: Professor Dr. Matthias Schmid