Calibrated Student t scores gave no observed mean loss gain in eight SEEG patients after contact order reversal with four recording pairs
We tested whether a fixed Student t response model reduced probability error after recorded stimulation contact order reversal. We used public STReEF SEEG data from OpenNeuro ds005448 version 1.0.0. Eight patients supplied four conditions and ten pulses per condition. Four selected bipolar recording pairs supplied 32 cosine coefficients in the 20 to 100 ms window. Five pulses fitted each reader. Five separate pulses supplied the test. Both response models received a confidence scale from fitting pulses alone. Mean test loss was 0.833562 bits per pulse for Student t and 0.826873 for Gaussian. The mean difference was +0.006689 bits. A descriptive 95 percent patient resampling interval was -0.019587 to +0.033459 bits. The exact sign-flip reference gave p = 0.59375. Raw Gaussian loss was 3.211361 bits. The calibrated comparison showed no observed Student t mean gain. A proof shows that the Student t transform changes score gaps but preserves class order. The package includes three figures, all test probabilities, fit arrays, source code, source receipts, and replay instructions. All 320 fresh raw ranges matched their saved hashes. An independent check matched the fits and probabilities. An offline replay matched all sixteen numerical outputs. This is an exploratory result in a reused clinical sample. The split is later within each condition. It is not a global time split. The study does not test a live brain interface or a brain-writing policy.
SEEG · intracranial EEG · electrical brain stimulation · probability calibration · log loss · Student t model · Gaussian model · STReEF
References
6- usesThe nearest half of SEEG pairs had a higher mean rate of repeatable waveforms across recorded contact order reversal in eight STReEF patients sci:sha256:399c1198c4184a27bd9be26923336885a19d4fdd7d9a5d49a4ca8f9cffac5cae
We used the inherited source metadata selection and public source retrieval recipes. The package preserves the pinned metadata and a replay that matches the 32 selected conditions.
- usesTwo calibration pulses gave lower mean four-condition accuracy than five at four SEEG pairs in eight STReEF patients after recorded contact order reversal sci:sha256:058302082013f6e5addde4db54df44a678225953923aa2f35d633ef3c5c03f68
We used its complete four-condition compact inputs, native examples, pinned source receipts, and direct raw retrieval helpers. We kept the five-pulse split and checked every raw range again. We added the Student t probability comparison and nested confidence fits.
- usesSelection for distinct response patterns had an uncertain mean accuracy gain at four SEEG pairs in eight STReEF patients sci:sha256:143d7f91f05a486692072b6c1e61b67efa46729f12681baae0430cec9016d9ad
We used its four-pair greedy log determinant rule, diagonal variance rule, eight cosine coefficients, and retrospective candidate-pair definitions. Each calibration fold refitted the pair rule without its held-out pulse.
- usesdoi:10.18112/openneuro.ds005448.v1.0.0
Source dataset, STReEF OpenNeuro ds005448 version 1.0.0; only the eight SEEG patients were analyzed.
- doi:10.1016/j.ynirp.2025.100274
- doi:10.1007/s12021-022-09567-6
Cited by
2 publicationsPublic catalog entries on Sciscape.
- usesStudent t scores reduced mean probability loss with simulated fourfold voltage gain in five SEEG tests at 100–500 ms
Uses its prior calibrated Student t comparison as the method basis. The new test uses repeated responses in ds004457 and simulated voltage gain, rather than contact order reversal in ds005448.
- usesFive-pulse products raised mean log loss in eight SEEG patients versus geometric means after recorded contact order reversal
This parent supplied the fixed eight-patient sample, feature arrays, source receipts, fitting code, and single-pulse confidence rule. The new study keeps those inputs and compares repeated-pulse probability pools. All 32 single-pulse reader settings matched the parent.