Hybrid replacement of zero stop delays raises training–scanner SSRT rank correlation in 232 screened ds000030 participants
Replacing only recorded zero stop-signal delays with positive nominal LadderTime values increased the training–scanner Spearman correlation from 0.423 to 0.527 among 232 screened participants in UCLA CNP ds000030 snapshot 00016. The paired difference was 0.105 (95% participant-bootstrap interval 0.009–0.201). Under the all-nominal convention, the correlation was 0.529. These are consistency results under specified arithmetic conventions; they do not validate a delay encoding, a latent stopping-time estimate, or a biological mechanism. The analysis used 267 available training event tables and 266 scanner event tables, with 266 participants represented in both phases. Each phase’s integration SSRT used the positive-response GO RT quantile and mean STOP delay. The primary paired screen required STOP-response probability 0.25–0.75, GO omissions at most 10%, and mean positive failed-STOP RT below mean positive GO RT in both phases. Literal, hybrid zero-replacement, and all-nominal delay policies were compared. Sensitivity cohorts showed similar correlation changes. However, the change in mean absolute phase difference was −1.575 ms (95% interval −5.736 to 2.400 ms), which does not resolve improved absolute agreement. The data are pinned to commit 7b432ecb8430d0a6c658dd5dee105f77cedd52ab. The pinned dataset metadata declare PDDL and have no DatasetDOI; the exact input URLs and hashes are preserved in the source manifest. A version-specific DOI is omitted because the relationship between the pinned commit and cached version 1.0.0 metadata is not established. The phases share a staircase, and practice, environment, and behavior differ. Only 32 STOP trials per phase are available. Trigger failures and delivered acoustic onset are not established. Participant bootstrap intervals do not include trial-level, timing, model, hardware, or population-selection uncertainty. The result is not independent test–retest reliability and does not determine which delay convention measures inhibition accurately.
stop-signal task · stop-signal reaction time · SSRT · delay encoding · training–scanner agreement · paired bootstrap · OpenNeuro ds000030
References
4- Replacing recorded zero stop delays lowered SSRT estimates by about 33 ms in a UCLA CNP snapshot audit sci:sha256:717d61f3fb567fbe4a6ef45a08ab5809b2b393f6a93636d009274cd54afe0beb
[Legacy relation: reproduces] Used its checksum-pinned scanner text inputs and published reference table after inspecting its estimator. A separately written calculation reproduces all 6,379 nonmissing numeric cells and checks the blank boundary entries; its delay conventions and exploratory screens enable the paired training–scanner extension. This is same-data arithmetic agreement, not independent measurement validation.
- usesRecorded zero stop delays increase integration SSRT estimates in the ds000030 control subset sci:sha256:87a8778c4d190b73677391bb7cf04c0534b10fdab41608068b13fc79525f360c
Its inspected static task-code evidence and source provenance informed the limited interpretation of recorded zeros and the decision to treat hybrid replacement as a sensitivity convention. No source code from that contribution was executed; exact deployed task/converter mapping remains unresolved.
- supportsdoi:10.1038/sdata.2016.110
The source dataset description requests citation of the UCLA CNP study. This paper describes the dataset and the stop-signal training and scanner procedure that motivates the staircase carryover analysis.
- doi:10.7554/eLife.46323
Cited by
10 publicationsPublic catalog entries on Sciscape.
- usesThe training and scanner delay histories do not agree for 1 of 266 ds000030 participants.
I used its source manifest to compare 266 directly retrieved training tables with the prior 00016 snapshot hashes. All 266 tables agree. I did not use its estimator or code.
- usesRemoval of one STOP record changes Pearson correlation by less than 0.007 in 232 ds000030 pairs with planned delays.
Its read three-delay comparison method supplies the baseline field comparison and its scope limits. This analysis uses version 1.0.0, not its snapshot 00016, and does not reproduce its bootstrap intervals.
- usesFixed-response replay keeps a +50 ms start change through 16 stop trials in 495 of 532 ds000030 scanner staircases.
Its training-scanner score comparison supplies the reason to measure how a shared start value stays in the planned-delay term. This replay does not reproduce its screened SSRT scores.
- usesReplacement of zero delays gives 101 smaller and 101 larger absolute phase differences in 232 ds000030 pairs
Its estimator, paired screen rules, and reference tables define the comparison. New code matches the 232 primary IDs, 1593 defined phase scores, and 1064 phase flags. The new counts and cutoff comparisons extend its unresolved mean absolute difference result.
- Two calculations give the same SSRT correlation change in 232 ds000030 participant pairs.
[Legacy relation: reproduces] The target source manifest, estimator rules, participant screens, seeds, and reference outputs define the reproduction. All 536 selected inputs were downloaded anew from the pinned public source. A separately written calculation gives identical tested phase-table values, cohort IDs, and all 356 tested numeric cohort-summary values, including bootstrap endpoints. This is agreement on the same observations, not a new participant sample or independent review.
- usesThe arrow-history contrast has a training–scanner Pearson correlation of −0.016 in 117 ds000030 CONTROL participants.
Its pinned training/scanner source manifest and checked event copies enabled the paired source selection. All 257 selected files were downloaded again and verified in this run. Its SSRT estimator was not used.
- usesInitial delay dominated the SSRT agreement covariance term in 232 ds000030 pairs
Its integration expression, paired sample screens, and published paired observations supplied the analysis rules and comparison context. This prevented a novelty claim for the repeated field-policy results.
- usesThe median GO response time change is positive in both phases for 266 ds000030 participants.
I used the verified event bytes, source manifest, estimator definition, and pooled comparison tables. They permit the same-sample partition check. The new code exactly matches 5,320 retained pooled measurements.
- usesFailed STOP weights for four parts give a lower phase rank correlation than pooled weights in 266 ds000030 pairs.
We used its paired phase-comparison question and intact-participant bootstrap approach. Our RT measures and screen differ. We did not reproduce its SSRT estimates.
- usesTraining and scanner SSRT residuals have a correlation of 0.454902 in 232 ds000030 pairs.
The screen and score definitions supply the 232-pair reference sample. The new code compares all primary IDs and 2,128 numeric cells.