Vigilance scores did not reduce ridge forecast error for later closed-eye fMRI motion in one person

We tested whether pre-session vigilance scores improved forecasts of mean motion in later closed-eye fMRI runs. We used OpenNeuro ds007328 version 1.0.0 and fixed ridge models. The primary sample had 111 sessions on 80 dates from one person. The first 30 dates supplied initial training. The test had 72 sessions on 50 later dates. Each fit used only earlier dates. Task scores increased measured log squared error by 1.89%. The error decrease was -0.003560. Its 95% seven-date block interval was [-0.008244, 0.000722]. This interval conditions on fixed forecasts and includes zero. An intercept-only reference performed better than both ridge models. Source checks found eight five-minute tasks and six nested task records that lacked published scores. The package preserves a recovery parser, all fits, source files, hashes, and a verified retrieval recipe. A new download produced identical analysis outputs. The result applies to these models and this single-person record. It does not establish an effect across people.

fMRI · framewise displacement · head motion · vigilance · ridge regression · forecasting · single-subject study