A fixed beta SEEG model had person median error above zero output in both test periods after simulated shaft loss

We tested a fixed spatial model for beta SEEG reconstruction after known simulated loss of a complete electrode shaft. The source was OpenNeuro ds006519 version 1.0.0, with 35 records from 21 people. The model used a 1.5 s baseline fit and a fixed ridge penalty of 0.1. No input shared a physical contact with the target. All 21 person median errors exceeded the zero-output reference in both test periods. The group median normalized errors were 1.586 in later baseline and 1.600 in marked post-stimulation windows. Zero output gives an error of 1. The mean paired person contrast was 0.0636, with a descriptive 95 percent bootstrap interval from -0.0661 to 0.2636. The exact sign test gave p = 0.383. A later penalty check showed that strong shrinkage reduced error toward 1 while output power fell. A small period contrast therefore did not establish useful reconstruction. The package preserves paired voltages, models, code, four figures, source hashes, and a retrieval recipe. Raw extraction and analysis replay checks passed. The evidence applies to these windows and models. It does not establish a causal stimulation effect or online BCI benefit.

SEEG · stereoelectroencephalography · beta band · signal reconstruction · ridge regression · brain-computer interface · electrode loss · intracranial EEG