27.05 11:00 - 12:00
USI East Campus, Room D0.02 |
|---|
| |
|---|
Abstract: Reducing the number of experimental units is a core principle of the 3Rs (Replace, Reduce, Refine) in animal research while maintaining statistical reliability. We propose a novel measure of evidentiary value per experimental unit by adapting diagnostic likelihood ratios and the diagnostic odds ratio to hypothesis testing and adjusting them for sample size. The resulting experimental unit information index (EUII) combines power, Type-I error, and sample size, with interpretations in both frequentist and Bayesian terms. We derive the EUII for simple tests and extend it to adaptive designs with early stopping for efficacy or futility. Applications to group-sequential and adaptive testing procedures, and a reanalysis of 2738 animal experiments with simulated interim analyses, demonstrate substantial sample-size savings. This is joint work with F. Balabdaoui, S. Fontana and S. Pawel.
Hosts: Prof. E. Wit and prof. D. Sulem | |
|---|
|
|---|
|
|
| | Leonhard Held earned his Ph.D. in Statistics at LMU Munich in 1997 under the supervision of Ludwig Fahrmeir. In 2000- 2006 he was Lecturer, Senior Lecturer and Associate Professor for Biostatistics at Imperial College London, Lancaster University and LMU Munich, respectively. In 2006 he joined the University of Zurich (UZH) as Full Professor. He delivered the Armitage Lecture at the University of Cambridge in 2015 on probabilistic forecasting of infectious disease spread. His current research interests are in reproducibility and replicability, clinical trials methodology, evidence synthesis and meta-science. He is founding Director of the UZH Center for Reproducible Science and Research Synthesis and Open Science Delegate of the University Board. 11:00 |
|---|
| |
|---|
|
|
|
|
