Alternating windows gave higher agreement of one-month posterior alpha changes than consecutive halves in 60 adults

We tested agreement of one-month posterior alpha log power changes in 60 adults from OpenNeuro ds004148 version 1.0.1. We calculated session 3 minus session 1 separately for two parts of each closed-eye record. Each part had 140 planned seconds. The score used 8 to 13 Hz, eight posterior channels, and a mean reference over 59 channels. Alternating windows gave ICC(A,1) 0.943. Consecutive halves gave 0.599. The paired difference was 0.344, with a 95% whole-person bootstrap interval of 0.176 to 0.515. The two parts had the same change sign in 55 of 60 people for alternating windows and 45 of 60 for consecutive halves. All seven planned descriptive checks gave a positive difference. They used the same dataset. The package preserves code, all 360 source spectra, raw retrieval hashes, the plan, results, and reproduction instructions. The offline replay reproduced all 13 analysis outputs. Four complete raw records rebuilt the stored arrays exactly. This test concerns recorded changes under fixed processing. Shared session effects can contribute to agreement. It cannot identify true neural change.

electroencephalography · alpha power · test-retest reliability · change agreement · intraclass correlation · split-half design · signal processing · bootstrap