Nominal 90 percent phase limits covered 35.3 percent of blocks of ten forecasts against an offline target in seven SEEG rest records

We tested phase error limits in seven public SEEG rest records from OpenNeuro ds006065, version 1.0.0. A fixed reader predicted an offline theta phase target 50 ms ahead. Each person had four selected channel pairs. Forecasts occurred every 100 ms in stored sample order. The test had 9632 blocks, with ten forecasts per pair and block. Nominal 90 percent limits fitted for single forecasts covered 89.9 percent of individual targets. They covered all ten targets in 35.3 percent of blocks. Limits fitted on a separate block period covered 92.0 percent of blocks. The mean full width increased from 249.0 to 337.5 degrees. A full circle has a width of 360 degrees. Block limits returned a full circle for 59.7 percent of forecasts. All seven people had higher block coverage. A rank proof gives coverage only under exchangeable calibration and test blocks. That assumption was not established for these ordered records. The result supports tests that report both sequence coverage and phase precision. It does not show a stimulation benefit. The package contains voltage, fixed models, a sequence tool, source hashes, code, results, three figures, and reproduction instructions. Full source checks, model refits, and exact analysis replay passed. The test data and earlier reader results were already public.

stereoelectroencephalography · theta phase · phase forecasting · prediction intervals · sequence coverage · conformal prediction · brain-computer interface