Under patient sign symmetry, theta load tests did not reject zero under two references at hippocampus-labelled SEEG contacts in 14 patients

We tested whether verbal memory load changes late-delay theta power under two signal references. The source was OpenNeuro ds004752, version 1.0.1. We used the first two public sessions from 14 patients with epilepsy. The 95 SEEG anchors had supplied hippocampus labels and next-contact neighbors. Both references used matched anchors and trials. Theta covered 4 to 8 Hz. The outcome was delay power relative to fixation power. The contrast was eight minus four letters. The equal-patient mean contrast was -0.051 dB under the supplied reference and -0.117 dB after bipolar subtraction. Their patient bootstrap 95 percent intervals were -0.877 to 0.797 dB and -0.806 to 0.538 dB. The paired reference difference was -0.066 dB, with an interval of -0.688 to 0.558 dB. All three planned exact sign-flip tests failed to reject zero after Holm correction. These tests assume independent patients and patient sign symmetry. The result does not establish equivalence or the absence of a memory load effect. Patient reference differences ranged from -2.759 to 2.785 dB. Only 44 neighbor contacts also had hippocampus labels. The small clinical sample, spatial coverage, fixed windows, and outcome definition limit wider claims. The package preserves 18,626 trial-reference feature rows, source code, fixed source hashes, metadata, patient results, figures, and reproduction instructions. All full source hashes passed. A separate reader and SciPy reproduced 128 powers in a small raw example. The raw audit covers one session. Offline analysis needs no full raw download.

verbal working memory · theta power · intracranial EEG · SEEG · signal reference · patient sign-flip test · OpenNeuro