Forecast magnitude ranked accurate 50 ms forecasts better than past sine amplitude against offline theta phase in seven patients' later SEEG rest data

We compared two confidence scores for the same calibrated 50 millisecond theta phase forecast. The test used 17 public SEEG rest files from seven patients, at 230 to 305 seconds. The target was phase from a specified offline 4 to 10 Hz filter and Hilbert transform. Mean AUC for error within 45 degrees was 0.6270 for forecast magnitude and 0.5513 for past sine amplitude. The patient mean difference was 0.0757; all seven patient and 17 file differences were positive. The descriptive patient bootstrap interval was 0.0609 to 0.0892. The exact two-sided sign-flip p-value was 0.015625, under independent patients and joint null sign symmetry. A fixed validation median threshold kept 51.11 percent of eligible forecasts and reduced mean error from 63.49 to 53.47 degrees. The sine threshold kept 50.32 percent and gave 59.83 degrees. Code, input excerpts, source versions, hashes, models, forecasts, checks, and rebuild commands are preserved. This is a later-sample test within a reused cohort. It does not establish biological phase truth or stimulation benefit.

theta phase · SEEG · phase forecasting · forecast confidence · signal quality · linear prediction · AUC