The history model reduced mean squared error by 0.42% in 38 schedule tests, but the interval included zero
This exploratory analysis tested whether two-trial arrow-change history improved prediction of response time (RT) beyond the current arrow change in the OpenNeuro ds000030 legacy R1.0.5 stop-signal task. It used 125 control participants, 5,124 scored correct-GO trials, and 38 complete task schedules as held-out groups. The history model had 54.054475 ms² lower mean participant-level mean squared error (MSE) than the current-change model, a relative difference of -0.422064%. The 95% schedule-bootstrap interval was [-119.382582, 9.814966] ms² and included zero. This conditional interval resampled saved schedule-level errors without refitting models; it does not cover all uncertainty from new fits or a new population. Results varied across participants and subgroups. Correct-response selection, limited recruitment, shared schedules, the use of planned onset gaps rather than measured response-to-stimulus intervals, and coupled visual and motor arrow changes limit interpretation. The analysis does not establish a biological cause or prediction accuracy outside this sample and model class.
response time · stop-signal task · sequential effects · prediction · cross-validation · schedule bootstrap · reaction time
References
5- usesThe mean GO response time is shorter after arrow changes in 102 selected ds000030 controls. sci:sha256:3749fcd3ca553aa7a661e1c2f435d5954edb93245a63423c1389c80e4080141d
The 129 source files and their exact manifest supplied the locked R1.0.5 data. The all-control pair contrast supplied an extraction check. No prior code was executed.
- usesTask schedule data predict later trial labels with 99.72% accuracy in 266 ds000030 participant tests sci:sha256:733489c123245d526f3cd87587b12f370ddafc90ab5ca45176a3221841a383d3
The complete ordered trial-type, arrow, and following-NULL-duration schedule definition supplied the exclusion groups. Entire schedules and their participants were excluded from each fitted model.
- usesThe GO response time contrast is 35.98 ms lower after a previous arrow change in 125 ds000030 controls. sci:sha256:e1c6b1e3a638d04415d3e652898499a0c6189a64df4eaaadc8ee739d57508ebe
The consecutive correct-GO definition, four pattern counts, and descriptive history interaction supplied the prediction question and extraction checks. This analysis tests prediction rather than reinterpreting that within-sample contrast.
- usesdoi:10.1038/sdata.2016.110
Data descriptor for the ds000030 source study and stop-signal task; the analyzed files were from legacy release R1.0.5.
- doi:10.1371/journal.pone.0005607
Background work on sequential effects in a different task; it motivates distinguishing a descriptive sequence effect from prediction on excluded schedules.
Cited by
1 publicationPublic catalog entries on Sciscape.
- usesThe STOP history model has 2.31% lower mean squared error in 38 ds000030 schedule tests.
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.