Sensorimotor brain–computer interface performance depends on signal-to-noise ratio but not connectivity of the mu rhythm in a multiverse analysis of longitudinal data

Citation: Nikolai Kapralov, Mina Jamshidi Idaji, Tilman Stephani, Alina Studenova, Carmen Vidaurre, Tomas Ros, Arno Villringer and Vadim Nikulin, “Sensorimotor brain–computer interface performance depends on signal-to-noise ratio but not connectivity of the mu rhythm in a multiverse analysis of longitudinal data,” J. Neural Eng. 21 056027, DOI 10.1088/1741-2552/ad7a24

• The paper is accessible through J Neuroengineering

Summary: This study presents a large-scale statistical benchmark that systematically compares 24 analysis pipelines in a multiverse framework to evaluate the robustness and reproducibility of predictors of BCI performance across up to 11 sessions in 62 participants. Using a longitudinal, session-wise design, it demonstrates that signal-to-noise ratio effects are consistently detected across pipelines, whereas other metrics are highly pipeline-dependent and lose significance under controlled comparisons.