A paired beta-power test gave an inconclusive result in 21 SEEG patients

We tested whether absolute beta-power changes across marked pre/post windows exceeded changes within the pre window. We used OpenNeuro ds006519, snapshot 1.0.0. The available sample had 35 sessions in 21 patients with negative motor responses to stimulation. We used equal 1.5-second power windows and gave each patient equal weight. The mean excess change was 0.154 dB. The exact two-sided patient sign-flip test gave p = 0.0957. This test assumes independent patient endpoints with symmetric distributions about zero under the null. It did not establish excess change at the 0.05 level. It also did not establish equal changes. A twofold beta-power rule flagged a mean of 16.1% of channels within baseline and 22.2% across pre/post windows. These are power-change flags, not artifact labels. Exploratory checks changed the result; a later post window gave p = 0.0481. The data have no exact train offsets or matched-gap sham controls. Thus, they cannot define amplifier recovery or a stimulation blanking time. The package contains seven figures, all feature tables, code, source metadata, source hashes, validation, and a retrieval recipe.

SEEG · beta power · brain-computer interfaces · electrical stimulation · power features · sign-flip test · source audit