Improved methods for analyzing multiple discrete tests

• Applicant: Professor Dr. Sebastian Döhler
• Subject Area: Mathematics
• Term: from 2020 to 2024
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 446840866

In modern scientific investigations, researchers are analyzing ever bigger data in order to discover interesting signals. This often involves testing thousands or millions of statistical hypotheses simultaneously, which increases the risk of making false discoveries (so called ‘type 1 errors’). Contemporary multiple testing research aims at developing theory and methods for controlling the extent of such false discoveries, even in very high-dimensional data, while still being able to reliably detect true signals. Most of the currently available methods have been developed in the setting of continuously distributed test statistics, which makes them conservative when discrete data such as counts or frequencies are analyzed. This project aims at closing the gap by developing theory and methods that use the information inherent in discrete tests more efficiently, yielding more powerful statistical procedures for analyzing multiple discrete tests.

• DFG Programme: Research Grants
• International Connection: France
• Cooperation Partner: Dr. Etienne Roquain