Separate scales for 16 slow modes increased mean relative SEEG forecast error at 20–500 ms in eight patients after contact order reversal

This test used public SEEG data from eight STReEF patients and 96 conditions. Five early pulses set each forecast. Five later pulses tested it after recorded stimulation contact order reversed. Separate scales for 16 slow DCT modes increased mean relative voltage error by 0.80% against one shared scale. The target was strict 20 < time < 500 ms. Seven patients had higher error. The descriptive 95% patient bootstrap interval for the increase was 0.24% to 1.23%. The primary sign-flip reference gave p = 0.03125 under independent and sign-symmetric patient differences. The score gave equal readout weight within conditions and equal condition and patient weight. With early power error weights, the mean increase was 0.074%, and its interval included zero. The sample was reused and exploratory. Time and contact order effects cannot be separated. The package supplies an oracle risk proof with explicit assumptions, four figures, code, exact compact inputs, all scores, and repeat instructions. All 960 raw range hashes and 147 metadata hashes matched public sources. A fresh raw rebuild gave exactly equal coefficients. Independent formula and waveform checks passed. A full numeric replay matched all 21 main output files. This test does not establish full-voltage or live interface performance.

stereoelectroencephalography · brain stimulation · voltage forecast · calibration · DCT · component scales · forecast error