AI-based biostatistical evaluation in biological sciences for non-statisticians

7&8 April 2025

Ludwig Hothorn, Leibniz Hannover University, DE

A major complication of the statistical evaluation by non-statisticians is the conversion of an evaluation strategy/model into executable R code. This can now be overcome by specific AI apps, such as Rtutor.ai. Based on experimental data, well organized in xls files, plain English textual instructions are converted into executable R code and the data is evaluated accordingly. The code can then be adapted/extended with certain R skills to generate graphics and tables directly suitable for a paper. The course focuses primarily on significance tests to demonstrate the treatment effects in oneway and multi-way designs with/without repeated measures-typical in molecular plant sciences and biological sciences. First, the appropriate arrangement of the data in xls files from a statistical point of view is explained in detail - this is already modeling. Secondly, some case studies from various experiments from UNIL and UNIGE groups as well as publications from Nature are used to teach the appropriate test/modeling strategy with the help of selected CRAN packages as plain text input. Subsequently, the improvement of the R-code for direct use in papers will be practiced. Finally, examples of overcoming, evaluated by the participants, will be explicitly discussed.

UNIL

12