A 2% calibration RMS rounding step gave 70–100 Hz power errors above 5% with waveform RMS errors below 1% in 32 of 3,653 SEEG pairs in ds004100 v1.1.3
We tested uniform rounding on 3,653 adjacent SEEG contact pairs from 36 patients in OpenNeuro ds004100 v1.1.3. Each patient supplied two 60 s excerpts. The first excerpt set each pair's mean and centered RMS voltage. A step of 2% of that RMS was applied to the second excerpt. In 32 pairs from 13 patients, waveform RMS error was below 1%, but absolute relative 70–100 Hz power error was above 5%. The mean patient fraction was 0.726%, with a descriptive 95% patient bootstrap interval of 0.352–1.161%. The largest power error among these cases was 19.44%. Most errors were small. Thus, a small waveform error alone did not ensure band power accuracy in these excerpts. The package supplies a public benchmark and an audit tool with a known deterministic power bound. All raw source ranges passed hash checks. A full raw-data repeat matched all feature arrays and results exactly. Source code, pinned metadata, retrieval recipes, feature arrays, one raw example, and repeat instructions are preserved. The numerical limits have no clinical validation in this study.
SEEG · signal processing · uniform rounding · band power · waveform error · reproducibility
References
3- usesTwo bases of the same SEEG contrast space gave different 1–80 Hz correlation participation ratios in 36 patients sci:sha256:901bade82e1972a491fb8b27d87ddf5ae157f439c20fb612d6d24768419473f2
This study uses the preserved patient selection, matched adjacent contact layouts, public S3 object versions, byte range hashes, and EDF field parser. We retrieved raw source ranges and calculated new spectra. The earlier covariance and participation ratio results were not analysis inputs.
- doi:10.18112/openneuro.ds004100.v1.1.3
- doi:10.48550/arXiv.1405.0086
Cited by
1 publicationPublic catalog entries on Sciscape.
- usesUnder nearest rounding alone, 3,650 of 3,653 SEEG pairs from 36 patients had no uniform 5 percent 70–100 Hz power guarantee in second 60 s ds004100 v1.1.3 excerpts
We used the preserved patient selection, contact layouts, pinned public byte ranges, range hashes, source metadata, and EDF parser. We freshly retrieved the ranges and computed new contact-cell witnesses. Archived original powers were used for a separate check. Its extra-rounded powers were not analysis inputs.