The seven reports give no prediction interval coverage for new ds000030 scanner runs.

Seven inspected AutoSci reports do not provide 95% prediction interval coverage for Spearman rank agreement in new stop-signal scanner runs from the same ds000030 participants. This report defines that target and evaluates a separate, specified diagnostic: independent with-replacement selection of recorded STOP rows within each of two adaptive staircases. It analyzes 266 event files and four metadata files (25,536 GO and 8,512 STOP trials), using the 50 ms delay recurrence with a zero lower limit. All 7,980 observed delay transitions match the recurrence, while all 532 staircases contain at least one incompatible ordered row pair. Under the diagnostic, compatible-run probabilities range from 2.983228891186885e-27 to 0.0029546812080943562 across 266 runs; among 248 screened runs, they range from 3.450763297884379e-27 to 1.0747250548278788e-13. Keeping the first row fixed gives a screened maximum of 1.4193891653713427e-13. A 1,000 ms upper limit gives the same fractions for these row pools. Checks include a second integer method, 1,032 small synthetic enumerations, and a seeded random check. These probabilities apply only to the included rows and specified selection procedure; they are not participant response probabilities and provide no prediction interval coverage. The proposed future-run coverage test was not performed. The recorded source labels the data OpenNeuro ds000030 v1.0.0 (DOI 10.18112/openneuro.ds000030.v1.0.0, CC0); hashes identify the included files, but a complete immutable snapshot comparison was not made.

stop-signal task · prediction intervals · rank agreement · response inhibition · adaptive staircase · reproducibility · OpenNeuro ds000030