Fixed 28-second ridge fits of hand acceleration increased mean later 4–8 Hz posterior EEG error during closed-eye rest in 25 ds003490 v1.1.0 Parkinson participants

We tested whether hand acceleration predicts simultaneous posterior EEG in later rest windows. The source was OpenNeuro ds003490 version 1.1.0, with two visits from each of 25 Parkinson participants. Each fixed model used 28 seconds of training data, three acceleration axes, and ridge factor 1. The primary test used closed eyes, a mean scalp reference, and frequencies from 4 to below 8 Hz. Mean later error increased by 7.33% relative to the training-mean EEG baseline. The 95% person bootstrap interval for the increase was 1.65% to 14.37%. Training error fell by 29.63%, and only 4 of 25 person scores improved in later windows. Ridge factor 10 gave a small uncertain reduction of 0.84%. The result applies to the fixed short fit and does not establish absence of shared signals or EEG artifacts. The package preserves phase information for 1,400 matched windows, model weights, predictions, errors, bootstrap draws, and source instructions. All 100 fresh raw ranges matched saved hashes. An independent solver checked 6,352 frequency fits. An offline rerun matched all 15 analysis files byte for byte.

electroencephalography · EEG · hand acceleration · Parkinson disease · complex Fourier values · ridge regression · prediction · resting state · cross-validation · measurement artifacts