High gamma differences within shafts lowered mean paired window loss on later trials in eight SEEG recordings

This study tested whether high gamma differences within electrode shafts add useful task-window information beyond beta and shaft means. It used 2,030 public SEEG trials from eight patients in OpenNeuro ds004473, version 1.0.1. Each model selected its penalty on earlier trials and was tested on later trials. Shared shaft scales made the reduced model a precise coefficient restriction of the full model. Mean loss fell from 0.4124 to 0.3551 nats, a gain of 0.0573 nats or 13.9 percent. Seven patients improved. The descriptive patient bootstrap interval was 0.0257 to 0.0917 nats. The exact sign-flip value was 0.0234375 under independent patients and joint null sign symmetry. All four fixed sensitivity checks kept positive mean gains. Fresh public raw bytes reproduced every input. A second complete analysis and an independent optimizer checked the results. The package preserves code, inputs, source versions, hashes, models, results, figures, and repeat instructions. This is a secondary analysis of reused recordings and test trials. It supports this fitting rule and paired-window test. It does not establish a neural source, new-patient prediction, or clinical benefit.

stereoelectroencephalography · high gamma · beta power · within-shaft contrasts · paired-window loss · ridge regression · time split · secondary analysis