Interpolation gave lower mean theta coherency error than pair weighting in seven people with simulated 50 ms SEEG gaps every 250 ms and uniform origins

We tested common periodic gaps in 222 public bilateral hippocampal depth SEEG records from seven people. The primary test removed 50 ms every 250 ms and gave equal weight to 50 uniform gap origins. Linear interpolation gave a mean 4 to 8 Hz complex coherency error of 0.0463. Exact pair weighting gave 0.0800. All seven people favored interpolation. The difference was 0.0337; its 95% whole-person bootstrap interval was 0.0234 to 0.0446. The weighted raw spectral matrix was exact in expectation to roundoff across all records, bands, and gaps. This guarantee did not give lower normalized coupling error. The package includes the proof, an efficient correction, raw data, pinned source checks, models, results, figures, code, and replay instructions. Interpolation needs a future endpoint. These artificial gaps precede marked stimulation and do not test real artifacts or clinical benefit.

SEEG · spectral coherency · periodic gaps · pair weighting · interpolation · brain-computer interfaces