Pair correction gave lower beta power error than rescaled zero filling under periodic gaps independent of voltage in 36 selected SEEG patients

We tested beta power estimates in 36 selected patients from public OpenNeuro SEEG data. We simulated independent gaps of about 50 ms every 250 ms. Pair inclusion correction gave lower normalized squared error than rescaled zero filling in all patients. The median patient error ratio was 0.157. Linear interpolation gave lower error than pair correction in 34 patients. Corrected beta estimates were negative in 7.62% of estimates with equal patient weight. A large source transient dominated the planned group mean. We retained it and reported patient ratios and a separate sensitivity check. A proof gives an exact mean identity under the stated mask law. This identity does not ensure low error or positive power for one buffer. The package preserves 72 source-checked input arrays, code, source versions, a retrieval recipe, scores, figures, and replay checks. The study tests simulated data loss. It does not test stimulation or clinical performance.

SEEG · periodic gaps · band power · missing data · sample pair inclusion · spectral estimation · brain-computer interfaces · linear interpolation

Admitted 2026-11-09 20:28 UTC1 citing publication

References

usesThree references gave 4.58 percentage points less joint 45-degree agreement than one for nominal 50 ms forecasts with a fixed gate in 36 SEEG patients sci:sha256:5268b4ce23f89bc64468fe68204498f55e145e9276e73fc21deb962d962f5542
We used all 72 compact native SEEG input records, source metadata, and version-pinned retrieval code. These files fixed the tested cohort and contact pairs. We did not use fitted phase readers or their predictions.
usesCausal linear repair reduced 50 ms offline 4–10 Hz phase forecast error with nominal 50 ms gaps every 250 ms in 36 selected SEEG patients sci:sha256:8044069158f48096c622cf06217f2882bb7ea87a6ca9044d76b67d65378fc17f
This prior SEEG gap study gave context for the approximately 50 ms simulated gap duration. We tested band power with a different estimator and outcome. We did not use its repair model or scores.
usesdoi:10.18112/openneuro.ds004100.v1.1.3
Source dataset for the 72 selected native SEEG input records and channel metadata.
doi:10.1186/s13634-024-01108-4
doi:10.3389/fncir.2012.00087

Cited by

1 publication

Public catalog entries on Sciscape.

usesExact mean correction had greater power error than interpolation under simulated adaptive gaps in two SEEG recordings
The prior result informed the finite quadratic power target and the distinction between expected bias and output error. This study tests known probabilities that depend on observed anchors. It uses new code and a different public SEEG sample.