The four ds000030 metadata files do not give a timing error limit for the tested screen.

A review of four metadata files for OpenNeuro ds000030 version 1.0.0 and the read task sections of its original data paper found no measured limit for the difference between GO and STOP mean timing errors. This finding is limited to those records and does not establish that no calibration record exists elsewhere. Using all 266 scanner stop-signal event files from the saved Git tree, the inherited responding GO minus responding STOP reaction-time rule passed in 252 files; the full screen, which also applies STOP-response and GO-omission rules, passed in 241. Removing each of 128 trials once in each file gave 34,048 separate removal states. Under an assumed per-response timing-error bound epsilon, the exact interval for the mean difference is [D - 2 epsilon, D + 2 epsilon]. At epsilon values of 0.01, 0.1, 0.5, and 1 ms, respectively, 2, 26, 120, and 180 removal states were ambiguous for the reaction-time rule. These are conditional arithmetic results, not measured errors, confidence intervals, or probabilities. All 266 original full-screen decisions remained fixed across the four tested limits. The work does not establish that actual timing errors changed a decision, and the tested metadata and task sections do not cover the full dataset record.

reaction time · stop-signal task · timing error · measurement calibration · OpenNeuro · ds000030 · sensitivity analysis · trial removal