Felix Schroeder

The 80% power lie
statmodeling.stat.columbia.edu/2026/02/19/t...

application for this phd closes on 10 Feb

Frontiers | The impact of cross-validation choices on pBCI classification metrics: lessons for transparent reporting Neuroadaptive technologies are a type of passive Brain-computer interface (pBCI) that aim to incorporate implicit user-state information into human-machine i...

Our results demonstrate how easily CV choices can distort conclusions, particularly for more complex models, and underscore the importance of transparency.

Thanks to Stephen Fairclough, Frederic Dehais, and Matt Richins for their guidance!

📄 Paper: doi.org/pt6h

#Neurotech #BCI #OpenScience #EEG

🔍 Takeaways for offline pBCI studies:
- Clearly document your data-splitting strategy — subtle choices can skew results
- If using blocks: keep them short, interleave conditions, and don’t split a block across train/test
- Trial-level randomisation is usually better to avoid time-based confounds

Differences in mental workload detection accuracy across various classifiers and cross-validation schemes. There is a trend that the more free parameters a model has, the greater the accuracy inflation in block-structure independent cross-validation schemes.

Cross-validation can seriously mislead passive Brain-Computer Interface (pBCI) evaluation.

🎉 New publication!

In our latest paper, we demonstrate how cross-validation choices can influence pBCI performance in mental workload detection and potentially mislead model comparisons.

(1/3)

AI, LLMs, and BS – Data Visualization with R Use R, ggplot2, and the principles of graphic design to create beautiful and truthful visualizations of data

It was probably this! datavizsp25.classes.andrewheiss.com/resource/ai-...