A centered 100 ms mean used future voltage after causal filtering in 10 SEEG recordings

We tested whether a centered power mean uses future voltage after a causal band filter. We used public SEEG data from OpenNeuro ds006519 version 1.0.0. A fixed edge rule selected 10 recordings from six people out of 35 public recordings. We changed only voltage at and after each detected edge. The causal 70 to 110 Hz band path with a centered 100 ms power mean changed the earlier feature in all 40 selected channels. The median person score was 0.433 baseline power units. Its descriptive 95% person bootstrap interval was 0.308 to 4.028. Fourteen channel scores exceeded one unit. The same causal band path with a past-only mean gave exactly unchanged earlier output. Both future changes and all three tested mean widths passed this causal control. The proof and an exact time-shift check show how to label the centered feature by its availability time. Detected edges are not known delivered pulse onsets. The test measures input dependence, not clean neural voltage or stimulation safety. The package preserves exact selected source voltage for all 35 records, source hashes and metadata, code, complete results, four figures, and reproduction instructions. Independent raw sample checks and an exact numeric repeat passed.

causal filtering · power features · stereoelectroencephalography · brain-computer interfaces · signal processing · future-input dependence