Fitted weights reduced the largest reference change variance on average in paired 20 s, 2–40 Hz SEEG samples from 36 patients

Removal of a shaft changes the global reference at all retained SEEG contacts. We tested fixed positive weights in 36 patients from OpenNeuro ds004100, version 1.1.3. Each patient supplied two 20 s samples with 2–40 Hz unit gain. The rule mixed uniform shaft weights with inverse standard deviation weights, in equal parts. Only the other file supplied the voltage data for fitting. Both file directions were tested. The primary loss was the largest reference change variance across single shaft removals. The geometric reduction against equal shaft weights was 32.4 percent. The 95 percent patient bootstrap interval was 28.3 to 36.3 percent, under independent patient sampling. All 36 patient mean gains and 70 of 72 file direction gains were positive. Top 10 percent link replacement changed by -0.22 percentage points, with interval -1.23 to +0.74. The link result did not show a clear stability gain. This is a secondary analysis of the same public cohort used in the cited AutoSci source. It does not establish a better clinical reference. The package preserves code, covariance inputs, pinned source retrieval, all scores, uncertainty draws, figures, and checks. Fresh downloads reproduced all 72 covariance inputs exactly. A second analysis reproduced all numerical outputs.

stereo-EEG · SEEG · global reference · shaft exclusion · reference variance · signal processing · epilepsy