Nominal 90 percent Gaussian voltage arcs gave a full circle in 95.20 percent of 50 ms theta forecasts against an offline reference in seven SEEG rest recordings

We tested whether early voltage statistics can give useful phase limits without current phase labels. We used public rest SEEG from seven people and 28 fixed bipolar pairs. The source was OpenNeuro ds006065 version 1.0.0. The target was a centered offline theta reference, 50 ms after each forecast cutoff. At the nominal 90 percent level, Gaussian voltage arcs covered 99.20 percent of later reference phases. However, 95.20 percent of outputs covered the full circle. Mean full width was 348.03 degrees. Only 1.25 percent of outputs had a half-width of 45 degrees or less. Coverage among these selected outputs was 83.90 percent, with equal weight per available pair and person. An extra phase calibration period gave 89.90 percent coverage and a mean full width of 255.30 degrees. The Gaussian proof requires true, stable moments and gives marginal coverage only. Fitted moments and selection for narrow outputs do not have that guarantee. This result limits the tested method for this offline benchmark. It does not establish live stimulation precision. The package preserves voltage, models, code, source hashes, results, and four figures. All seven full public source files and exact derived traces passed fresh checks. A complete repeat analysis matched all numeric results.

SEEG · theta phase · phase forecasting · Gaussian uncertainty · signal processing · intracranial EEG · OpenNeuro