The 95% interval included zero for one-month agreement of adjacent four-second 8-13 Hz log power correlation in 60 ds004148 adults with closed eyes

We tested a time order score in OpenNeuro ds004148 version 1.0.1. The main test used 60 adults with closed eyes at visits one month apart. Each record gave 70 planned four-second windows from seconds 10 to 290. We integrated 8-13 Hz power over eight posterior channels under a 59-channel average reference. The score was atanh(q). Here q was Pearson correlation between adjacent accepted log power values. Rejected windows broke adjacency. Absolute agreement ICC(A,1) was 0.229. Its 95% person bootstrap interval was -0.073 to 0.543. The one-sided p value for 10,000 random order draws was 0.0374. The planned positive rule required an interval above zero and p no greater than 0.05. That rule was not met. Mean adjacent q was 0.272, with a 95% interval of 0.228 to 0.318. No random order draw reached that mean. Mean power ICC on the same people was 0.831. The paired power-minus-order ICC difference was 0.602, with an interval of 0.285 to 0.892. All five processing check agreement intervals included zero. The package preserves inputs, source hashes, code, the local plan, all planned settings, uncertainty draws, figures, and reproduction commands. Independent calculations and three fresh raw source replays passed. The full offline repeat matched all saved numerical outputs. The random order reference is conditional on accepted values and masks. It does not identify a neural cause. The evidence concerns this cohort and method.

EEG · alpha power · test-retest reliability · intraclass correlation · time order · OpenNeuro ds004148