Occupied FRandR time gave lower mean overlap than counts across dates in seven SEEG patients under independent uniform tie choices

We tested whether occupied event time improves channel list agreement across sleep recording dates. We used supplied FRandR marks from public OpenNeuro ds003498 version 1.0.1. The primary set had 135 records of 300 seconds from seven patients with depth-only implants. Each record selected ceil(10% of the channel count) channels. Overlap was expected Dice under independent uniform cutoff tie choices. We gave each date pair equal weight within patients and each patient equal weight. Mean overlap was 0.64324 for event counts and 0.63803 for occupied time. The change was -0.00521, with a descriptive 95% patient bootstrap interval of -0.08107 to 0.05810. The exact two-sided sign-reversal test gave p=0.9375 under independent patients and symmetric null contrasts. Four patient contrasts were positive, two were negative, and one was zero. The result does not show a clear mean gain or establish equivalence. Union time equaled summed duration in every observed channel map. We prove that occupied time stays unchanged when intervals are split or repeated with the same union. The proof assumes fixed coordinates, record limits, channel universes, and tie rules. The package holds source files, hashes, maps, code, proofs, results, figures, and reproduction commands. This is an annotation comparison in a reused cohort. Repeated timestamps and a SEEG/ECOG label conflict limit interpretation. It does not test clinical accuracy.

high-frequency oscillations · intracranial EEG · SEEG · epilepsy · event duration · occupied time · channel map agreement · reproducibility