A constrained fit reduced mean 7 Hz filter phase error against a guard applied after fitting in nine SEEG records with simulated mixed gaps and known initial states

We tested whether fitting under a fixed state map norm limit reduced error during missing SEEG voltage. We used nine public rest records, four fixed bipolar pairs per record, and four completed gap lengths. Both fitted methods used the same 0.95 limit, features, early data, and ridge penalties. On simulated mixed gaps, mean 7 Hz filter phase error was 50.97 degrees for the constrained fit and 65.24 degrees for a scalar guard applied after fitting. All nine record means improved. Mean normalized state squared error fell from 1.3350 to 0.6147. The constrained fit had 0.295 degrees more phase error on burst timing. An added comparison with singular value clipping also favored the constrained fit on mixed timing. The test assumed known initial states. It does not establish true neural phase accuracy or a stimulation benefit. The package preserves voltage, source receipts, code, fit certificates, five figures, and reproduction instructions. Fresh source prefixes matched all selected samples. A separate fit and replay reproduced every main fit and score file exactly.

SEEG · signal processing · phase estimation · constrained ridge regression · missing voltage · state estimation