The order test did not detect excess 20 to 500 ms SEEG waveform change in eight patients

We tested whether early and late pulse means differ more than random splits of the same pulses. We used public SEEG data from OpenNeuro ds004696 version 1.0.1. The data have eight patients, 86 stimulation sites, and 886 local bipolar readouts. Each site used its first five and final five good pulses. The primary window was 20 to 500 ms after stimulation. The squared waveform difference was compared with an exact random split expectation. The mean excess was 2.7%. The patient bootstrap 95% interval was -3.0% to 10.2%. The two-sided patient sign-flip p value was 0.641. That test assumes independent patients and a symmetric null distribution. It did not detect excess change. It does not show that pulse order has no effect. A 20 to 100 ms secondary check gave 11.7% excess and an unadjusted p value of 0.039. This result needs a separate confirmatory study. The package preserves the proof, code, exact input products, source hashes, results, and three figures. All imported hashes passed, and the numerical analysis reproduced exactly. An added source download did not complete; source waveform validation is a checked parent-study record. The method can support stimulation-response calibration for brain interfaces. No online interface performance or stimulation safety was tested.

stereoelectroencephalography · electrical stimulation · pulse order · waveform change · random split expectation · brain-computer interfaces