Biometric Methods

Biometric Methods: Model-based analysis of reproducibility in biomedical research

The reproducibility crisis, now widely recognized in the scientific community, involves the unnecessary use of experimental animals in some cases. Therefore, such non-reproducible research results must be identified quickly and, as far as possible, avoided in the future: Wrong results lead to wrong new research approaches and thus to unnecessary animal experiments. Therefore, one focus of research at the Bf3R is to find out which methodological and institutional changes in research and the publication process can lead to better reproducibility and efficiency and thus to a reduction of animal experiments.

The statistical methods prevailing in biomedical research – null hypothesis significance tests – were identified a root cause for the reproducibility crisis. New statistical approaches proposed by the statistical community to overcome this problem are therefore investigated to make the research process more efficient. New mathematical models are being developed for this purpose.

The statistical planning of animal experiments serves to ensure the validity of the corresponding experiments and, at the same time, to reduce the number of animals required for this purpose to the necessary minimum. Moreover, as part of the experiment, reproducibility can be checked with statistical methods and thus contribute to the gain of reliable data.

Further reading:

Steinfath, M., Vogl, S., Violet, N., Schwarz, F., Mielke, H., Selhorst, T., Greiner, M., & Schönfelder, G. (2018). Simple changes of individual studies can improve the reproducibility of the biomedical scientific process as a whole. PLOS ONE, 13(9), e0202762. https://doi.org/10.1371/journal.pone.0202762

Rudeck, J; Vogl, S; Banneke, S; Schoenfelder, G; Lewejohann, L (2020): Repeatability analysis improves the reliability of behavioral data. PLoS ONE, 15(4): e0230900. https://doi.org/10.1371/journal.pone.0230900