Exact normal limits reduced log width and gave 85.7 percent mean coverage for five-pulse batches at 83 SEEG sites in eight people

We tested limits for five recorded responses at 83 stimulation sites in eight people with epilepsy. The public SEEG data are from OpenNeuro ds004696 version 1.0.1. Each site used five early pulses to fit limits and five later pulses to test them. The fixed readouts used two to nine remote electrode shafts. The primary signal was log time-centered bipolar RMS from 20 to 500 ms. An exact normal batch calculation accounts for the shared fit mean and standard deviation. It assumes stable, independent normal log sizes over time within each shaft. A spatial union bound does not require independent shafts. The exact rule reduced mean relative log width by 8.40% against the Bonferroni batch rule. Mean complete-batch coverage across people was 85.7% versus 91.3%. The rules covered 74 versus 77 of 83 site batches. The descriptive 95% interval for exact coverage was 73.3% to 95.2%. It includes the 90% model target. The normal probability target averages over new fit and test samples. It is not a promise for each fitted site. This work supplies a numerical tool and a fixed test of an established prediction method. It does not test a live brain interface. The package preserves compact exact trial products, three figures, results, source hashes, and provenance. GPL-licensed code remains in the local archive and is omitted from SCI, so the SCI package alone cannot replay the analysis. A fresh full source rebuild for one person agreed in log RMS to 3.6e-15. The original clean analysis replay matched all result files by hash.

prediction intervals · simultaneous coverage · stereoelectroencephalography · SEEG · electrical stimulation · brain-computer interface · root mean square voltage · normal model