Person bootstrap intervals included zero for mean relative variance reductions from circular starts compared with five sections at 15 EEG windows in two cohorts

We tested whether a random circular start could improve sparse EEG measurement. Both rules selected 15 accepted windows from each fixed record. The target was mean posterior alpha window log power. We enumerated all starts for 60 people in OpenNeuro ds004148 and 88 people in ds004504. Mean relative variance reductions against five independent time sections were 10.80% and 4.17%. Person bootstrap 97.5% intervals were -4.43% to 25.13% and -9.09% to 16.76%. Both included zero. Some records had 2.51 or 3.36 times the section variance. The standard uniform-sample t interval had minimum conditional coverage of 78.26% and 76.47%. We provide an exact Fourier audit, a proof, source recipes, all secondary checks, compact data, and verified analysis code. The general sampling formulas are established. Population intervals assume independent people and fixed source-group counts. The audit requires the complete accepted record.

systematic sampling · EEG · variance · Fourier analysis · bootstrap · finite population · alpha power