In 8 SEEG subjects, the 70–150 Hz key-press map showed no detectable drift beyond 30-trial noise within one session (p = 0.34)

Question: during one SEEG session, does the channel map of the key-press effect change with time more than sampling noise explains? Data: archived per-trial bipolar log power (70–150 Hz and 13–30 Hz) from sci 04683a2a, OpenNeuro ds004473, 8 subjects, 132–298 trials, rest window vs key-press window. The SHA-256 hashes of the files are equal to the archived hashes. Method (plan hashed before analysis): for each 30-trial block we calculated a paired effect d for each channel. We compared the map of block A (trials 71–100) with the adjacent block (41–70) and with the last 30 trials (Pearson r, Fisher z). Primary test: exact one-sided sign-flip test over subjects on D = z(A,Far) − z(A,Near). Result: D was not negative (mean −0.042, median +0.022, 4 of 8 negative, p = 0.34). The median r with the last block (0.44) was near the median r of random disjoint 30-trial sets (0.41), so sampling noise, not time, limited the map similarity. The 13–30 Hz map gave the same result (p = 0.22). Map similarity did not decrease with the time gap (median Spearman rho 0.00). Raw shrinkage-LDA weights repeated poorly (median r 0.27). Haufe patterns repeated better in 6 of 8 subjects (median r 0.42), but this difference was not significant (p = 0.16). One subject (sub-5) had a gradual map change (rho −0.94). Limits: offline, cued windows, one session, 8 epilepsy patients, and noisy 30-trial maps; a null result does not prove zero drift. The package contains the code, features, plan, results, figure, and instructions to reproduce the analysis.

stereo-EEG · intracranial EEG · motor brain-computer interface · high gamma · effect map · sampling noise · within-session stability · LDA · Haufe patterns · OpenNeuro ds004473