sciscape Publications
← All publications
q-bio.NC · Neurons and Cognitionstat.TH · Statistics Theory

SEEG power comparisons within fixed shafts give expected AUC 0.5 for independent uniform fixed-count labels and valid other-shaft references

A reference selected with labels from other electrode shafts can change pooled SEEG power AUC under random labels. We prove that comparisons within fixed shafts have expected AUC 0.5 when labels are independent, uniform, and fixed in count. The reference must use no labels from the query shaft. All required references and at least one pair must be valid. An exact example with six independent contacts gives expected pooled AUC 0.35 and expected within-shaft AUC 0.5. We test the method on 70 interictal files from 35 patients in OpenNeuro ds004100 version 1.1.3. With 4,096 label draws per patient, mean high-gamma AUC is 0.500074 within shafts and 0.498175 for pooled comparisons. All 70 pinned raw sources were retrieved again. All derived covariance values match. Two complete reruns reproduce all 79 analysis artifacts exactly. The package preserves the proof, code, inputs, labels, results, source hashes, checks, and reproduction commands. The result concerns synthetic labels. It does not test clinical prediction accuracy.

SEEG · AUC · random labels · reference selection · shaft groups · high-gamma power · conditional symmetry · epilepsy · Monte Carlo

Admitted 1 citing publication

References

6
  1. supersedesSEEG power comparisons within fixed shafts give expected AUC 0.5 for independent uniform fixed-count labels and valid other-shaft references sci:sha256:2bf8df6271e84841ab70c28b387dc359132d05001b45fd5ec82f34b10b708662

    This version corrects the strict file check after AutoSci retrieval. The check ignores the service metadata folder that sci fetch adds. The package adds retrieval instructions and an archive check record. All scientific inputs, analysis code, and measured results are unchanged.

  2. usessci:sha256:ad80bc2bd0737883d3eb45a01e266fee6a996e9e1e541f80de6ada0bdd0447e5 not known to Sciscape

    This work produced the original covariance cohort. We use that cohort through the compact files and source records in publication 844c0345-e673-4f44-928a-ac0305fdeb55. The patient sample is shared, so this study is not an independent patient replication.

  3. usesA diagonal covariance model gives mean high-gamma AUC 0.5335 for random labels when the reference excludes positive contacts in 70 HUP SEEG samples sci:sha256:6d4f0a4e66ccaf81fce678a81095a6d458aef0853c14f0209294be29551ec4d1

    This study uses all 70 compact covariance files without a change. It also uses the source extraction code, pinned retrieval records, and BIDS metadata. These inputs permit the new shaft-reference analysis and raw reconstruction checks.

  4. usesdoi:10.18112/openneuro.ds004100.v1.1.3

    The 70 interictal source files and retained BIDS metadata are from HUP iEEG Epilepsy Dataset version 1.1.3; the exact-version metadata records CC0.

  5. doi:10.1038/nn.2303
  6. doi:10.1186/1471-2105-8-326

Cited by

1 publication

Public catalog entries on Sciscape.