F limits covered 89.8 percent of later three-coefficient SEEG responses after five fit pulses at 86 selected sites in eight patients

We tested joint limits for recorded brain stimulation responses using public SEEG data from OpenNeuro ds004696, version 1.0.1. Five pulses fitted each site model; 612 later pulses at 86 selected sites in eight patients were tested. Three fixed cosine coefficients formed each response vector, and patients had equal weight after averaging sites. Gaussian limits covered 52.5% of later vectors. Finite-sample F limits covered 89.8% (descriptive 95% interval 85.1–94.1%), with 26.08 times the Gaussian volume. A Student t box covered 90.5% at 40.2% of the F volume. With four coefficients, the F volume ratio rose to 13,186.35. The benchmark shows the cost of covariance uncertainty with few fit pulses. Readout selection, patient differences, and incomplete waveform coverage limit transfer. The exact F law requires independent normal vectors and a fixed readout; these conditions were not established in this cohort. The evidence supports further tests of response limits for brain-computer interfaces, but does not establish a clinical alarm. The SCI package preserves inputs, results, figures, source records, and the full scientific text; GPL analysis and retrieval scripts are omitted, so the published payload cannot run the original replay.

stereoelectroencephalography · brain-computer interface · prediction limits · finite-sample F distribution · covariance uncertainty · electrical brain stimulation · OpenNeuro ds004696