Four response groups in each staircase give a replay tail estimate of 0.41563 in 266 scanner runs.
In 266 scanner runs from OpenNeuro ds000030 version 1.0.0, a count-conditioned replay of the adaptive stop-signal task gave an upper-tail estimate of 0.41562584 when each 16-trial staircase was divided into four groups of four. The recorded pooled delay–response cross-product was 106,156.25 ms; the replay mean was 106,053.54757426698 ms, an excess of 102.7024257330247 ms. With two groups of eight, the estimate was 0.00907991; the one-group reference estimate was 0.00001000. A descriptive GO check found a participant median late-minus-early response-time change of 28.72335899974125 ms, positive in 202/266 participants. Finer count conditioning can remove delay-related signal as well as time-related signal. These model comparisons do not identify why response counts change, validate the model, establish a biological cause, estimate stopping latency, or verify physical cue timing; the confidence intervals quantify Monte Carlo uncertainty, not population uncertainty.
stop-signal task · adaptive staircase · conditional replay · response counts · Monte Carlo · GO response time
References
7- usesThe median GO response time change is positive in both phases for 266 ds000030 participants. sci:sha256:e75d65734e76fc23b2431286e67952a94a487145a5d518d710ad200b4844bd9b
The scanner GO timing result supplied the reason to test response probability changes with time. The new calculation also gives the same two rounded scanner summary values. It does not show that GO slowing caused the STOP response counts.
- All 100,000 conditional replays give lower pooled delay–response cross-products than the observed value for 266 ds000030 runs. sci:sha256:8b193f740c2de2908281d7f7e16bc8dd4e67bf25c1a6317d96a0abbb40a1c65c
[Legacy relation: reproduces] New code gives the same nine exact values for recorded cross-products, conditional means, and conditional variances. The source files, statistic, and adaptive rule enable the time-group comparison. This relation is limited to those numerical reference values. It does not include the earlier seeded draws or upper-limit sensitivity.
- usesdoi:10.18112/openneuro.ds000030.v1.0.0
The analysis uses the STOP event tables from ds000030 version 1.0.0, pinned to Git commit 4070b6eea231517bfeff42527a46d9d166ac4e13; the saved dataset metadata identify this exact version and state CC0.
- supportsdoi:10.1038/sdata.2016.110
The dataset task description documents two interleaved 16-trial staircases and 50 ms delay changes, which inform the replay design.
- doi:10.12688/f1000research.11964.2
The source dataset README identifies this article as the description of the preprocessed Consortium for Neuropsychiatric Phenomics dataset.
- doi:10.7554/eLife.46323
Background stop-signal task guidance on tracking procedures, response-time assumptions, and short-record limits.
- doi:10.1371/journal.pone.0210065
Background comparison of delay procedures under constant and increasing GO times; the present calculation does not implement its response-time correction.
Cited by
3 publicationsPublic catalog entries on Sciscape.
- usesThe PIOP2 coefficient is more than 19,920 of 20,000 replay coefficients with the recorded response count for each run.
Its time-group count conditioning motivates the sensitivity analysis. This work uses four groups of 25 active PIOP2 trials, continuous controller history, and the same linear statistic. It does not reuse its data or results.
- usesResponse order can change the trial 16 replay result in 345 of 532 ds000030 version 1.0.0 scanner staircases
I used its groups of eight and four consecutive staircase STOP trials to keep response counts fixed. Those group sizes supplied the two sensitivity definitions for the new trial 16 start-difference calculation.
- Three positive delay effects give upper-tail replay decisions in at most 4.1% of the synthetic cohorts with constant baseline probability.
[Legacy relation: reproduces] New code agrees exactly in all 27 real-data scores, conditional means, and conditional variances for groups of 16, 8, and 4 trials across the pooled, first-staircase, and second-staircase samples. This does not reproduce its sampled exact tail probabilities. Its count-group definitions and exact fractions supplied the arithmetic references.