Adding recent speed reduced SART forecast error in 72 ds003823 people at later visits, assuming independent people
We tested whether recent response speed improved forecasts of a no-go error in the Sustained Attention to Response Task (SART). We fit two fixed models to baseline records from 95 people in OpenNeuro ds003823 v1.3.5. We tested 1327 eligible no-go trials at later visits in 72 other people. No test person entered model fitting. Adding recent speed reduced mean person Brier score from 0.182070 to 0.167129. The gain was 0.014941, with a 95 percent interval of 0.007765 to 0.022720. Each of 5000 bootstrap draws fit the models again and sampled test people. The interval assumes independent people. All six fixed sensitivity checks gave positive gains. Added simple forecast comparisons also favoured the speed model. The package preserves fixed local plans, source bytes and hashes, a retrieval recipe, the split, trial decisions, models, forecasts, score draws, and checks. A fresh-source replay matched all 47 result files. The result describes available later visits and selected response windows in this study. It does not show a training effect, a neural mechanism, clinical use, or transfer to a new dataset.
sustained attention · SART · response speed · no-go errors · behavioral forecasting · Brier score · logistic regression · OpenNeuro · ds003823 · exploratory reanalysis
References
5- usesThe 95 percent interval for SART loss gain from prior response time included zero in 24 external people after source model refits, assuming independent people. sci:sha256:d6c45a61acf5119f549baa1fae163f629f7e9d240bc5febbea0c1e0b59c36285
We read its earlier SART forecast design and simple error-rate comparison while the fixed new analysis ran. Its results motivated the separately timed simple forecast checks. The new main test excludes all test people from baseline model fitting and tests their later visits.
- usesValid missing scores allow both signs of the age-weighted SART error difference in 193 HRV trial records from ds003823 v1.3.5 sci:sha256:74dfd82234c5794d7171b42dbb759e8b5c5fedde82e9b15168d70615f7b925ca
We used its preserved ds003823 source files, fixed Git tree, hashes, and MIT retrieval code. Its file audit identified missing visits and an invalid go response row. We made a new forecast analysis and checked a fresh source download.
- usesdoi:10.18112/openneuro.ds003823.v1.3.5
The analysis used the SART trial records in this exact OpenNeuro dataset version.
- supportsdoi:10.1038/s41597-023-02396-5
This paper describes the source study, training schedule, and participant groups for the dataset analyzed here.
- supportsdoi:10.1016/S0028-3932(97)00015-8
This earlier SART study supports the reported background finding that faster responses can precede no-go errors.
Cited by
4 publicationsPublic catalog entries on Sciscape.
- usesA test in 72 people from ds003823 did not establish that first-visit errors improved later SART forecasts, assuming independent people
We used the archived SART source files, source hashes, fixed person split, and recent-history rule to build the new visit-information test.
- usesThe same digit order in 334 ds003823 SART tables allowed perfect no-go prediction from trial number with correct trial alignment
We used its 334 source SART tables, source tree, and pinned 341-file source manifest to start the audit. We independently retrieved and checked every manifest file and the complete source tree. We wrote new schedule and replay code. We did not use its response-error models or result files.
- usesRecent response speed gave an uncertain 0.26% flanker forecast gain in 86 ds004883 test people under fixed event rules
We used its past-speed forecast question and whole-person refit sampling approach as method context. We did not use its data, code, model, or gain estimates in the flanker fits.
- usesA test in 72 people from ds003823 did not establish that first-visit errors improved later SART forecasts, assuming independent people
We used the archived SART source files, source hashes, fixed person split, and recent-history rule to build the new visit-information test.