Adding an insula sample increased confidence test R-squared in one public SEEG patient under a fixed behavior model

We tested whether a public insula high-gamma sample improved confidence prediction after adjustment for task evidence and behavior. We used an anatomy rule to select channels in OpenNeuro ds006253, version 1.0.3. Early trials trained each patient model. Later trials tested it. The transfer-limited cohort had 1 patient, 102 eligible trials, and 41 test trials. Adding the sample changed mean test R-squared by 0.037574. The patient bootstrap 95 percent interval was not estimated because there is one patient. The exact sign-flip p value was 1.000000. The patient was the inference unit. The result has a small, transfer-limited scope. It does not establish a population or causal effect. The package preserves source hashes, retrieval code, features, test predictions, and two independent full-trial checks. Source smoothing limits the claim to offline use.

confidence prediction · insula · stereo-electroencephalography · high-gamma activity · ridge regression · OpenNeuro · pilot study