A sign flip test did not detect higher direction agreement after stimulation in seven SEEG patients. Source correction was fixed, and the test assumed independent patients and null sign symmetry.

We tested whether the stronger measured response direction repeats more often after stimulation than before stimulation. We used 190 reciprocal site pairs from seven epilepsy patients in OpenNeuro ds004696 version 1.0.1. Source correction used all trials. We kept this correction fixed. For each pair, we tested all 63,504 joint choices of separate five-pulse halves from ten pulses in each direction. We averaged powers from two destination contacts. The 15 to 500 ms agreement was 74.75 percent after stimulation and 77.54 percent before stimulation. Each patient had equal weight. The mean change was -2.79 percentage points. The conditional 95 percent patient bootstrap interval was -26.36 to +14.32 points. The exact two-sided patient sign flip probability was 0.90625. The test assumes independent patients and null sign symmetry. The wide interval does not establish an absent effect. Changes in contact choice changed the mean contrast sign. Two pairs had full agreement in early and late windows, but low agreement in the combined window. Opposite short-window direction contrasts nearly cancel. These pairs share a stimulation site. All numerical checks passed. A fresh source rebuild matched every saved array for one patient. We preserve data, code, results, figures, source versions, hashes, and instructions. This score measures repeatability of recorded power contrasts. It does not establish true biological direction.

stereo-EEG · electrical stimulation · response direction · repeatability · sign-flip test · OpenNeuro