T16 — Comparing methods using real and simulated data: challenges and best practices
Day, Time, Duration
Thursday afternoon, 2 pm‑6 pm, 4h
Language offered
English
Short Description
The course provides an overview of the methodology of comparative studies with a focus on statistical methods, but at the interface with other GMDS-relevant fields such as machine learning and bioinformatics. The target audience consists of readers/users of comparative studies as well as methodological junior scientists* who wish to design and conduct comparative studies themselves. In particular, the following topics will be covered: 1. Design of a comparative study (of methods), influence on results and danger of cherry-picking and non-replicable methodological research results 2. “Neutral comparative studies” as an ideal, difficulties and partial solutions 3. Reporting of comparative studies 4. Simulation studies in methodological research: design, practical relevance, neutrality, reporting 5. Data-driven simulation studies in applications as decision support for choice of methods Planned are course units in “lecture style” as well as short hands-on sessions in R on the above topics. Literature: M. Herrmann, P. Probst, R. Hornung, V. Jurinovic, A.-L. Boulesteix, 2021. large-scale benchmark study of survival prediction methods.
Organizer
Moritz Herrmann
Institution
Institute for Medical Information Processing, Biometry, and Epidemiology, LMU Munich, München
Contakt
moritzherr.uni [at] googlemail.com
boulesteix [at] ibe.med.uni-muenchen
c.niessl [at] ibe.med.uni-muenchen.de
Additional Speaker
Christina Nießl
Institution
Institute for Medical Information Processing, Biometry, and Epidemiology, LMU Munich, München