Constant convex mean19/Fz/Cz/Pz mixtures gave joint wPLI limits more than 0.01 below separate limits for three pairs in 32/88 ds004504 v1.0.9 records at 10 Hz

We tested ordinary sample weighted phase lag index (wPLI) in 88 processed EEG records from OpenNeuro ds004504 version 1.0.9. One constant convex mixture of mean19, Fz, Cz, and Pz had to serve F3/F4, C3/C4, and P3/P4 in each record. The separate limit U was the smallest of three separate pair maxima. The joint limit J was the highest minimum score under one mixture. At 10 Hz, U minus J was above 0.01 in 32 records (36.4 percent). The mean gap was 0.01731. Its 95 percent person-bootstrap interval was 0.01147 to 0.02402. Five records had U at least 0.20 and J below 0.20. We give a joint solver, an exact feasibility rule, attaining weights, and numerical upper exclusion witnesses. The largest joint bound width was below 3.0e-7 in all 440 planned record-setting cases. We checked all 3,520 joint witnesses, all separate upper vectors, exact examples, and an independent full linear program. A complete repeat agreed within 1e-9. Two fresh source replays matched the stored Fourier arrays exactly. The package preserves the full analysis input, source hashes, code, results, and instructions. The result concerns fixed sample scores in this reference family. It does not establish a brain connection or a disease effect. The cohort bootstrap holds recorded signals fixed and assumes independent people.

EEG · weighted phase lag index · wPLI · reference choice · convex optimization · connectivity estimation · feasibility · numerical certificates · OpenNeuro ds004504