Mean FRandR map agreement was lower than the independent timing model mean in 60-second sleep SEEG samples from eight patients

We tested top-channel map agreement in public sleep SEEG event marks from OpenNeuro ds003498 version 1.0.1. The final subset had eight patients with depth electrodes and 136 five-minute records. FRandR means a fast ripple that overlaps a ripple. We compared the first and last 60-second clips. We selected the top 20 percent of channel pairs, with a ceiling rule and equal weights for ties. The timing model fixed each full-record channel count and assigned event times independently and uniformly. Mean agreement was 0.6656 in the data and 0.7101 in 9,999 model draws. The deficit was 0.0445, with a conditional lower-tail p value of 0.0047. The 95 percent patient bootstrap interval was -0.0181 to 0.1335. Omitting patient 07, who had one eligible record, reduced the deficit to 0.0046 and gave p = 0.313. The result does not establish a general patient effect. A source audit found that the original ninth patient had depth and strip electrodes. We corrected the SEEG subset after the original result and preserved the plan and all original results. This is a new analysis of a reused cohort. The package contains 655 source files, source hashes, a source audit, code, all model draws, results, figures, and reproduction instructions. A separate calculation matched every observed score. A clean offline repeat matched all 20 core output files.

high-frequency oscillations · sleep SEEG · channel map agreement · conditional timing model · epilepsy