Frontal alpha asymmetry increased observed BDI prediction error in a fixed ridge test across two rest runs from 114 adults in OpenNeuro ds003478 v1.1.0

We tested whether frontal alpha asymmetry improves prediction of Beck Depression Inventory (BDI) scores. We used OpenNeuro ds003478 version 1.1.0. Fixed rules retained 114 adults. For each test person, we trained on the first rest run from the other 113 people. We tested on the second run from that person. The base ridge model used age, sex, and mean F3/F4 log alpha power. The added model also used F4 minus F3 log alpha power. We fixed the penalty before score analysis. Root mean squared error rose from 10.715 to 10.886 BDI points after asymmetry was added. Base minus added mean squared error was -3.694 squared BDI points. A 5,000-draw person bootstrap refit interval was [-15.143, 0.418]. The interval includes zero. All seven planned model checks also had negative observed gains. The result concerns these features, this fixed model, and this selected young sample. Population and clinical use require separate tests. The package preserves code, 243 source-linked feature records, individual predictions, all bootstrap draws, and reproduction instructions. Two new full raw record replays, all input hashes, an independent solver, and a complete repeated analysis passed.

frontal alpha asymmetry · EEG · BDI · ridge regression · person-held-out prediction · OpenNeuro ds003478