Beta scores selected from OFF have a negative swap mean in five screened Parkinson EEG pairs

We tested a beta frequency selection rule in OpenNeuro ds002778 version 1.0.5. The rule selects the largest OFF score and reports ON minus OFF at that frequency. A proof shows that selection repeated after each pair swap gives a nonpositive swap mean for finite score vectors on common candidates, with a fixed tie rule. The fixed EEG screen retained five of 13 eligible pairs. All 32 paired label assignments gave a mean of -0.189302 dB. A selector based on the mean of the two score vectors gave a zero swap mean. Separate windows changed the primary mean by +0.273320 dB. This change was zero with C3 and C4 alone. All seven planned settings are reported. Independent MNE and FFT checks passed for all 30 source recordings. A repeat analysis gave identical bytes for all 65 numerical files. The package preserves code, proof, spectra, screen values, source hashes, all resampling output, and a raw retrieval recipe. The small selected sample and remaining artifacts limit biological interpretation. The swap center does not identify a true medication effect.

EEG · frequency selection · paired comparisons · selection bias · Parkinson disease · signal processing · randomization · spectral analysis