Fixation power gave no clear recall prediction gain with ridge regression in later trials from seven patients with reported SEEG recordings

We tested whether fixation power predicts later recall after controls for trial order and past recall. We used public OpenNeuro ds006136 version 1.0.1. The main test used seven Utah patients whose recording files report SEEG counts. The source paper reports depth electrodes in these patients. Channel tables use EEG, and physical array types remain uncertain. We fit ridge regression on earlier trials and tested 846 later trials. Each patient had equal weight. Fixation beta and high gamma power reduced mean squared error by 0.001204 squared objects. This was 0.22 percent of control error. The 95 percent patient bootstrap interval was -0.007178 to 0.009409. The two-sided patient sign-flip test p value was 0.8125. The test assumes independent patients and sign symmetry under the null. The test did not show a clear prediction gain for these fixed models. Source cleaning before release limits live use. The package contains code, predictions, model fits, source hashes, and retrieval instructions. Independent checks and a repeated analysis reproduced the results.

fixation power · object recall · stereoelectroencephalography · ridge regression · memory · neural signal · forward prediction · OpenNeuro