A joint exchangeability test did not detect excess 20–500 ms waveform change across 86 SEEG sites in eight sessions

We tested the largest first-to-final waveform change at 86 stimulation sites in eight public SEEG sessions. The data are from OpenNeuro ds004696 version 1.0.1. Each site used its first five and final five good pulses. The exact global maximum test gave p = 0.456381 for 20–500 ms. The null requires joint exchangeability within each patient and independence between patients. This result does not establish stable neural responses. Ten pulses give 126 distinct balanced squared-distance assignments and a minimum exact p value of 1/126. Thus, no patient test can pass Holm correction across eight patient nulls at 0.05 with this design. The first feasible even pulse counts are twelve for eight separate tests and sixteen for 86 separate tests. These are resolution bounds, not power estimates. A descriptive 20–100 ms maximum gave p = 0.039683 for one global test. Its six-check Bonferroni reference was 0.238095. This is an exploratory reuse of the same cohort. The package contains a reusable exact maximum tool, five figures, all assignment scores, inputs, source hashes, and reproduction instructions. A separate run reproduced 58 numerical output files. Fresh source rebuilds checked two sessions; fresh metadata checks covered all 52 files.

SEEG · electrical stimulation · waveform separation · exchangeability · exact test · maximum statistic · multiple testing · OpenNeuro