One beta sign per link gave higher mean test scores than frequency signs in 19 SEEG records

We compared two phase sign rules in 19 public SEEG records from OpenNeuro ds003876, version 1.0.2. Both rules used the same adjacent contact differences and selected links. The first minute trained the signs. The next minute supplied the test values. The beta band contained 34 bins from 13 to 29.5 Hz. One rule used one sign per link. The other rule used one sign per frequency bin. The pooled rule gave a mean signed imaginary coherency score of 0.03151. The detailed rule gave 0.02766. The mean pooled gain was 0.00385. The 95 percent patient bootstrap interval was 0.00203 to 0.00559. Pooling gave a higher score in 17 of 19 records. An exact two-sided sign-flip test gave p = 0.000896. That test assumes independent patients and symmetric null contrasts. The detailed rule had a higher training score, as required by the score formula. All fixed checks retained the mean direction. This study reuses a published cohort and adds a new paired comparison. It is not an independent replication or a clinical test. All 19 public signal ranges rebuilt the archived Fourier inputs exactly. The package contains inputs, source pins, code, separate calculation checks, and reproduction instructions.

SEEG · stereoelectroencephalography · imaginary coherency · phase sign · beta band · signal processing · temporal transfer