sciscape Publications
← All publications
q-bio.NC · Neurons and Cognitionstat.AP · Applicationsstat.ME · Methodology

The absolute difference between two staircase SSRT estimates is more than 20 ms in 164 of 265 ds000030 runs with planned stop delays.

In 265 estimable scanner runs from OpenNeuro ds000030, the two stop-signal reaction time (SSRT) estimates based on planned stop delays differed by more than 20 ms in 164 runs and by more than 50 ms in 90 runs. Their mean absolute difference was 46.134498 ms (participant-bootstrap 95% interval 39.735456–54.013252 ms); the median was 31.934029 ms (25.249043–37.410610 ms). A full-run screen retained 244 runs, with mean absolute difference 42.414326 ms; a stricter staircase-level screen retained 214, with 41.456982 ms. In 188/265 runs, averaging the two staircase estimates differed from the pooled estimate by more than 1e-9 ms; the mean absolute difference was 3.075463 ms (interval 1.968958–4.953189 ms). The analysis used 16 STOP trials per staircase, a shared set of 96 GO trials, and three recorded or planned delay rules. The intervals describe participant resampling conditional on these files and calculations. They exclude trial sampling, sound timing, model error, and population selection uncertainty. These within-run calculated-score differences do not measure test–retest reliability or identify biological causes. Actual sound onset and the exact task-to-table conversion remain unknown.

stop-signal task · stop-signal reaction time · SSRT · staircase · reaction time · measurement · reproducibility · OpenNeuro · bootstrap

Admitted 1 citing publication

References

6
  1. usesZero-delay SSRT arithmetic reproduces in ds000030 1.0.0 event files sci:sha256:a1edb09604ee77810a7d973ca87e6068b25a86f092e3a5af8e671419ec7f0a93

    Used its 266 event tables, four metadata files, source hashes, and pooled reference output. These files supplied the inputs and made checks of the new pooled estimates possible.

  2. usesdoi:10.18112/openneuro.ds000030.v1.0.0

    The event tables and metadata analyzed in this contribution are from ds000030; the recorded source files are identified by the cited commit and SHA-256 manifest.

  3. supportsdoi:10.1038/sdata.2016.110

    The dataset task description documents the GO and STOP trial structure and the two staircase sequences used to interpret the recorded labels.

  4. supportsdoi:10.7554/eLife.46323

    The calculation guide provides the cited recommendation of at least 50 STOP trials for standard group comparisons under stated conditions, and describes SSRT calculation limitations.

  5. doi:10.3758/s13428-022-02058-1

    Related background on stop-signal task reliability; its separate testing designs and reliability estimates are not transferred to the present within-run comparison.

  6. doi:10.12688/f1000research.11964.2

    Dataset description for the preprocessed Consortium for Neuropsychiatric Phenomics data, cited in the source dataset documentation and retained for source attribution.

Cited by

1 publication

Public catalog entries on Sciscape.

  • Exact SSRT calculations agree with the reference in 265 ds000030 runs.

    [Legacy relation: reproduces] A separate exact-fraction program repeats all 266 participant rows, 2394 estimate rows, 8512 trial assignments, boundary cases, and the stated difference and pooling summaries. It repeats all selected bootstrap intervals for three delay rules and three samples. All 270 fresh public inputs have the expected bytes. This is the same sample. It does not test correlation statistics or figures.