The primary 95% interval for current confidence log loss gain included zero in 17 MetaRDK patients under fixed ridge models with later test trials

We tested whether current confidence improved recorded correctness prediction beyond task variables and previous confidence. Separate ridge models trained on earlier trials and scored later trials from 17 MetaRDK patients. The primary penalty was 10. Mean log loss gain was 0.010231 nats. The 95% person bootstrap interval was -0.002272 to 0.020945 nats. It included zero. Three planned secondary checks had intervals above zero. These checks shared source data and did not change the primary rule. The finding concerns stored outcomes after a confidence report in one patient visit. The package preserves source data, fixed retrieval references, new code, all ten comparisons, models, and figures. A separate optimizer checked all 342 fits. A fresh download reproduced 52 numerical files and 90 bootstrap arrays exactly.

metacognition · confidence · perceptual decision-making · logistic regression · log loss · temporal validation · epilepsy