One STOP trial removal changes inverse quantile SSRT by more than 20 ms in 8 of 121 ds000030 controls with recorded delays.

In a fixed-record sensitivity analysis of 121 controls from ds000030 snapshot 1.0.0, we removed each of 32 STOP trials in turn and recalculated inverse-quantile stop-signal reaction time (SSRT). Under recorded delays, the median participant-level maximum absolute change was 10.417646 ms; 8/121 controls exceeded 20 ms, and the largest change was 31.697000355611237 ms. Replacing zero delays with planned delays and using planned delays throughout gave 2/121 controls above 20 ms for each rule. These descriptive results depend on the inverse quantile and delay rules. The removals keep later adaptive-staircase records fixed and do not estimate sampling uncertainty, repeat-session reliability, clinical risk, or biological causality. The 20 ms threshold is descriptive, not clinical.

stop-signal task · stop-signal reaction time · SSRT · trial removal · sensitivity analysis · inverse quantile · reaction time · ds000030

Admitted 2026-11-09 14:15 UTC3 citing publications

References

Zero-coded stop delays increased estimated SSRT by 29.8 ms in 121 controls from OpenNeuro ds000030 v1.0.0 sci:sha256:33ad52286b43a0ca5455f0ff67dd9a848d126c4015232e23a36e01e178b08211
[Legacy relation: reproduces] The new code reproduces the source hashes, 121-control selection, and 1,752 available numeric baseline cells from the target. It then calculates one STOP row removal. It does not reproduce the target bootstrap interval.
usesdoi:10.18112/openneuro.ds000030.v1.0.0
The event tables analyzed here are from ds000030 snapshot 1.0.0; the saved dataset metadata identifies this exact snapshot DOI and states CC0.
supportsdoi:10.1038/sdata.2016.110
The source paper supplies the task design context of 96 GO and 32 STOP trials per run.
supportsdoi:10.7554/eLife.46323
The consensus guide supplies the cited STOP-trial recommendation and methodological context; this analysis does not test that recommendation.
doi:10.3389/fpsyg.2012.00037
Background on measurement and reliability of response inhibition; its reliability analyses and data were not reproduced here.

Cited by

3 publications

Public catalog entries on Sciscape.

4,856 of 7,260 ds000030 participant pairs keep their SSRT sequence with two delay rules and a maximum of one STOP row removal from each run.
[Legacy relation: reproduces] The new code agrees on the 121-person selection and 16,380 A/B score cells. It compares full-table inverse scores and removal scores for two quantiles. It does not compare all target outputs.
Two GO rank rules give positive and negative SSRT changes after the same GO row removal in each of 121 ds000030 controls.
[Legacy relation: reproduces] The source subset, manifest, sample rules, and reference tables enable this calculation. New exact code agrees with 378 initial upper-rule scores and 24,192 post-STOP upper/linear scores for 126 controls, under three delay rules. The largest difference is 1.7053025658242404e-13 ms. The same 121-person primary sample and reported threshold counts agree. This relation does not claim reproduction of every target field.
Removal of one STOP trial gives a zero delay difference in 24 of 84 ds000030 control runs.
[Legacy relation: reproduces] Recalculated all 8,064 STOP-row removal records for the 126 shared runs and two quantile methods. All 104,832 compared cells agree exactly. The 121-run absolute-score maxima and descriptive 20 ms counts agree. The observations are shared.