Past response-time variation gives an uncertain gain in error forecasts under fixed ridge models and stored trial assignments in 23 people in ds007216 v1.1.0
We tested past response-time variation as an added predictor of recorded mountain responses. The source was OpenNeuro ds007216 v1.1.0. We used fixed ridge logistic models. We kept each of 23 test people out of training. The primary test used 3,035 targets from 43 runs. A 20-trial history ended three trials before each target. The variation model reduced mean person log loss by 0.00711 beyond speed, recent errors, and task time. The 95% range from 2000 person resamples with model refits was -0.00168 to 0.01968. Average ranking within a person did not improve. The package preserves seven planned checks, source files, hashes, code, predictions, and resampling records. The result depends on stored response assignments. It does not establish a warning system during the task.
sustained attention · GradCPT · response-time variation · error forecasting · ridge logistic regression · person-level resampling · OpenNeuro ds007216
References
5- usesThe GradCPT event dictionary labels 1,199 stored mountain responses as correct in ds007216 version 1.1.0 sci:sha256:fbe126385e4c18577db9ef2367c805be3090846048bee9f40d625fe9154e8f4f
The source audit supplied the fixed event files, source manifest, retrieval code, and joint cue-response rule. We used them to test error forecasts from past response times. We independently checked source counts and feature values.
- usesdoi:10.18112/openneuro.ds007216.v1.1.0
The study analyzed the 43 functional GradCPT event tables and dictionaries selected from ds007216 v1.1.0; the source files and hashes are preserved.
- doi:10.64898/2026.02.04.703882
- doi:10.3758/s13414-012-0413-x
- doi:10.1016/j.neuroimage.2018.01.002
Cited by
3 publicationsPublic catalog entries on Sciscape.
- usesThe 95% person bootstrap interval included zero for guarded city response variation asymmetry in 23 ds007216 v1.1.0 people under fixed assignments.
The prior uncertain forecast gain helped define the new question about before and after descriptive contrasts. No fitted model or prediction from that study was used.
- usesAll 84 gradCPT main blocks have the same stored image code sequence in ds006040 version 1.0.1
The fixed ridge model and person bootstrap with model refits informed our analysis plan. We test prior-run responses at identical stored trial positions in a different dataset. We did not reuse its code or data.
- usesPast target response history gives no clear gain in error forecasts under fixed ridge models in 15 test people in ds006768 v1.1.0
Its guarded history, person weighting, and model refit uncertainty test informed this design. We tested mean target latency and past errors in a different moving dot task. We used separate fixed training and test groups. No data or code were copied from that study.