The fixed transfer model gave no clear later recall prediction gain after early calibration in nine SEEG patients

The fixed transfer model used theta and high gamma. It gave no clear prediction gain on later recall trials. We used OpenNeuro ds005059 release 1.0.6 and reused verified derivatives from 18 SEEG patients. Nine development patients supplied shared slopes. Nine evaluation patients supplied early labels for intercept calibration. Their 492 later trials supplied scores only. The patient mean Brier gain was 0.00003135. The 95% patient bootstrap interval was [-0.00007980, 0.00015799]. The exact sign-flip p-value was 0.640625. These measures assume independent patients and are conditional on the fitted source models and calibration samples. The package preserves inputs, source pins, code, model artifacts, predictions, and reproduction instructions. This is a new analysis of a reused cohort. It does not establish equivalence or a physiological null effect.

SEEG · associative memory · recall prediction · model transfer · patient calibration