Pair selection with test labels raised random label accuracy to 60.3 percent in eight SEEG recordings
We tested whether pair selection can create false SEEG decoding when it uses test labels. We used OpenNeuro ds004473 version 1.0.2 and 200 random trial label swaps per person. Each classifier used the same 32 fit trials, eight bipolar pairs, and three power bands. Selection with fit and test labels gave 60.34 percent mean balanced accuracy. Selection with fit and separate calibration labels gave 49.78 percent. Both rules ranked pairs with 72 trials. The paired mean increase was 10.56 percentage points. Its descriptive 95 percent patient bootstrap interval was 9.37 to 11.55 points. All eight recordings had an increase. A rule that excludes the independent random test labels has conditional expected accuracy of one half. The package preserves features, raw source checks, all models and predictions, code, figures, a proof, and reproduction instructions. This is an offline method audit for reading BCI studies. It does not establish live BCI performance.
SEEG · brain-computer interface · data leakage · pair selection · random labels · balanced accuracy · cross-validation · circular analysis
References
4- usesEight SEEG pairs classify pre-response and rest windows at 70.7% mean balanced accuracy in later trials from eight recordings sci:sha256:c74068384c7c056974c8568aab987cf5a8eee9979a622c22efa1b67a911e77fd
We copied all eight ds004473 version 1.0.2 feature files, trial records, source metadata, and extraction code. These inputs enabled the matched-size pair selection audit. The parent used earlier-trial selection; our invalid test-label rule is a new audit condition.
- usesdoi:10.18112/openneuro.ds004473.v1.0.2
The analysis used the eight recordings in OpenNeuro ds004473 version 1.0.2.
- doi:10.1038/nn.2303
- doi:10.1088/1741-2552%2Facae0a
This is the source study accompanying the SEEG dataset.
Cited by
1 publicationPublic catalog entries on Sciscape.
- supersedesPair selection with test labels raised random label accuracy to 60.3 percent in eight SEEG recordings
This version fixes package verification after sci fetch. The hash check skips the separate .autosci-publication metadata folder and checks every archived study file. The study evidence is the same.