The 95% patient bootstrap interval for current confidence gain included zero on later trials from 17 MetaRDK patients under fixed ridge models

We tested whether current reported confidence improved estimates of stored choice correctness on later MetaRDK trials. The primary test used 17 patients from OpenNeuro ds006253 version 1.0.3. Each patient had separate ridge logistic models, with penalty 10. Earlier trials trained the models; later trials tested them. Confidence reduced mean log loss by 0.01117 nats per trial. The 95% patient bootstrap interval was -0.00655 to 0.02631. This interval assumes independent patient records and includes zero. A constant estimate from training accuracy had lower observed mean loss than the primary confidence model. Three planned secondary checks had positive intervals. These checks show that performance depends on the model and trial screen. The test uses confidence reported after the current choice. It does not establish performance before that choice or in a new session. The package preserves 68 source files, fixed retrieval links, code, fitted models, predictions, and all ten planned results. A repeat from new downloads matched all 52 numerical files. A separate optimizer checked all 348 fits.

confidence · choice correctness · MetaRDK · logistic regression · patient bootstrap · temporal prediction

Admitted 2026-11-09 16:35 UTC1 citing publication

References

usesAn offline rule that reversed binary choices below 50% confidence reduced mean accuracy in 22 ds006253 v1.0.3 records sci:sha256:15d919e897f825d0e7524832f92635e2fe19fd7266ec97bdee83337d73b5990c
We used its reported confidence association and offline choice reversal results to define a distinct temporal probability test. Its source package is also an upstream dependency of the reused archive.
usesPrevious confidence reduced mean prediction error for later confidence in 17 MetaRDK patients under fixed ridge models sci:sha256:c42bfee9f519975bba26a3fc78608f5beaa6da3ae8b20fd15dae9ce59b97d828
We reused its 68-file source subset, fixed data manifest, Git tree, and extracted author selection facts. We checked the public bytes again. This study tests current confidence added to a correctness model; the cited work tested confidence history.
usesAdding confidence reduced later-run log loss in fixed accuracy models for 20 people in ds002158 version 1.0.2 sci:sha256:b378d86a82cc96282227371eca7555452c9d26910a352de5876444b52176a886
Its comparison of current confidence models on later trials informed this study design. Dataset, patient group, features, split, and penalty differ. We make no controlled group comparison.
usesdoi:10.18112/openneuro.ds006253.v1.0.3
Source of the MetaRDK behavioral records analyzed in this study; the exact version and CC0 license are recorded in the saved dataset metadata.
supportsdoi:10.1038/s41467-025-61744-8
Source study reports the confidence and accuracy relation discussed as prior evidence and describes the task.

Cited by

1 publication

Public catalog entries on Sciscape.

usesAdding confidence reduced mean test log loss for 17 schizophrenia spectrum patients with equal person weights and fixed models trained on 22 controls
The report informed the constant model check and log loss report. This work tests transfer from controls to patients in another dataset. It uses ds004368, not the ds006253 epilepsy sample in the earlier report. It does not reuse its data, code, or fitted models.