The STOP history model has 2.31% lower mean squared error in 38 ds000030 schedule tests.
In a comparison of later correct GO reaction times from 125 CONTROL participants in the recorded ds000030 R1.0.5 event files, a linear model that included STOP history had 2.3121810209953906% lower participant-mean squared prediction error than a base model across 38 schedule-excluded tests. The models produced 8,768 predictions. The history model had lower participant MSE for 80 of 125 people and lower mean MSE in 34 of 38 schedules. Its MSE difference was -386.4133407515892 ms² (conditional schedule-bootstrap 95% interval [-499.5458619331418, -276.9828790268982]); its MAE difference was -1.462177103016515 ms (interval [-1.9583585031145596, -1.0184108092656559]). These intervals use fixed predictions and do not include refitting uncertainty. Test participants contributed early baseline calibration. The result applies to correct responses in these records and does not establish causality, prediction without personal calibration, prediction of errors or omissions, or generalization to another dataset. The analysis did not estimate stopping latency or brain activity.
stop-signal task · reaction time · STOP history · prediction error · schedule exclusion · cross-validation · GO response · neuropsychology
References
4- usesThe history model reduced mean squared error by 0.42% in 38 schedule tests, but the interval included zero sci:sha256:1c845199641968a4d97377ab5867acc56178e48a5dc56411eb562c2b88cd770b
Its early-32-position personal baseline, equal-participant fit weights, complete-schedule exclusions, and conditional schedule bootstrap supplied the evaluation design. New code tests STOP history on a different later correct-GO sample. Its arrow-history numerical result was not repeated.
- usesMean temporal half-run SSRT is at least 10 ms from full-run SSRT in 11 of 102 selected ds000030 runs. sci:sha256:36fb5f06074b933615cf2bf61f06b40034ed65f51a017cb9f571c32bdf68176e
Its 129-file ds000030 R1.0.5 source manifest and fixed 102-run eligibility table supplied the public source lock and a secondary score subset. All 129 source files were downloaded again and verified. Its SSRT calculation was not repeated.
- usesMean GO reaction time is higher at the first three GO trials after STOP in 125 ds000030 control runs. sci:sha256:fa8f124bcd2975628f040bd689fe94803e19689ebc065949085d6f58efa4d31e
Its nearest-STOP rule for the previous three active positions supplied the history inputs and prediction question. This run tests excluded-schedule prediction error with new code. It does not repeat the earlier descriptive means or claim causality.
- doi:10.3389/fpsyg.2013.00649
Cited by
1 publicationPublic catalog entries on Sciscape.
- usesStarting delay reduces scheduled SSRT leave-one-out mean squared error by 7.28% in 107 screened ds000030 control pairs.
Its complete-schedule exclusion design motivates the scanner trial-pattern holdout check. Here patterns use ordered trial type, arrow direction, and ladder number for 256 rows. The target is scanner SSRT in version 1.0.0; its later-GO target, history features, and R1.0.5 data are not reused.