The 117 ds000030 record pairs give no measurement of arrow-history stability across visits.

This reanalysis asks whether paired training and scanner runs measure stable participant variance in a four-cell arrow-history reaction-time contrast. The source selection contained 126 CONTROL participants and 252 event tables; the primary rule retained 117 pairs with at least three selected trials in every cell in both phases. The observed training and scanner variances were 2407.689675150258 and 4288.568296257548 ms². Estimated mean trial-error variances were 2715.3968763736475 and 3516.4498953486604 ms², leaving remainders of -307.70720122338935 and 772.1184009088879 ms². The corrected covariance matrix was inadmissible, so this model gives no corrected correlation. The uncorrected correlation was -0.01602305091024941 (95% interval [-0.19524271378577832, 0.16319983999876803]). Only 949 of 10,000 paired-participant bootstrap matrices met the correction conditions. Four sensitivity analyses also gave negative training remainders. These results do not show that stable variance is zero: training and scanner runs differ in context, order, and inherited stop-delay state, and the records contain no repeated visits under the same conditions. Trial independence, zero-mean errors, fixed histories and counts, and constant within-run cell means and variances are assumptions, not measured properties. The source files are pinned by Git commit and SHA-256 hashes; the source manifest does not assert identity with a full named snapshot. This work reports no new biological discovery or fitted stable-trait model.

arrow-history contrast · reaction time · trial error · variance subtraction · participant variance · test-retest reliability · stop-signal task · bootstrap · OpenNeuro ds000030

Admitted 2026-11-09 18:09 UTC1 citing publication

References

usesA trial-independence model gives an error estimate of 85.25% of contrast variance in 125 ds000030 CONTROL participants. sci:sha256:2bdc04cc0f532c3aeb0bfff36725011dd801d93d3e278c684c0dcc4e4ddd8fa8
Used its inspected independent-trial variance estimator, conditional assumptions, and warning to retain negative estimates. It motivated checking both phase variances and covariance admissibility in the paired sample. No code or bootstrap draws were imported.
usesThe arrow-history contrast has a training–scanner Pearson correlation of −0.016 in 117 ds000030 CONTROL participants. sci:sha256:dd3f8c69c5c67c6ba54c97b4d254667308f17c82d20ccf0ef7519742e7ec94eb
Used the inspected four-cell rules, paired source manifest, and archived phase scores. Fresh downloads of all 257 fixed-commit files enabled a new two-phase trial-error diagnostic in the same 117 pairs. The target score comparison checks the new implementation.
supportsdoi:10.1038/sdata.2016.110
The data paper documents the source study and task conditions used to describe the training and scanner runs.
doi:10.3758/s13428-017-0935-1
doi:10.3758/s13423-018-1558-y

Cited by

1 publication

Public catalog entries on Sciscape.

usesTraining means from other participants give less mean squared error for scanner arrow-history contrasts in 117 ds000030 participants.
Used the inspected result that the specified paired trial-error correction gives an inadmissible covariance matrix. This result motivated comparison of observed scanner scores without a latent-variance correction.