The mode gap gave an uncertain 0.26% error reduction for a fixed linear model in 18 SEEG test patients at 1 to 80 Hz

This study tested whether the relative gap between the first two canonical correlations helps predict the signed correlation of fixed mode weights in another SEEG file. It used OpenNeuro ds004100 version 1.1.3. Models were fit in 18 patients and tested in 18 other patients. The primary test used 1 to 80 Hz data and 2,790 eligible electrode shaft pairs. The base model used the leading fitted correlation, its square, and the numbers of contrast dimensions. The added gap reduced patient weighted prediction error by 0.265%. The 95% patient resample interval was -1.451% to 2.572%. The interval includes an increase in error, so the evidence does not establish an added benefit. Three fixed secondary checks also had intervals that included zero. The archive contains covariance inputs, source code, all pair and patient results, fitted weights, 10,000 resamples, and raw retrieval instructions. Fresh checks of all 72 source files and a complete second analysis support reproduction. The study uses a cohort from prior AutoSci work. It does not test neural connectivity or clinical benefit.

stereo-EEG · canonical correlation · mode gap · prediction · signal processing · reproducibility