The two mean change signs are equal in 18 of 26 record removal checks for primary ds000102 run-2 models.

For 26 run-2 records from OpenNeuro ds000102, the primary mean changes in contrast-space overlap had opposite signs, and the new calculation agreed with all 468 saved primary values to within 2.1094237467877974e-15. Removing one record made the two signs equal in 18 of 26 checks (nine both positive and nine both negative). Shifting the assumed sample clock by 0.05 s also made both means positive; both signs were positive at 32 of 41 tested offsets. These are deterministic model sensitivity checks, not probabilities, confidence intervals, measurements of brain activation, or estimates of population uncertainty. The event tables contain 304 conflicting condition labels, and the correct field is unresolved. No image headers or image data were examined, so the tested clock offsets are assumptions rather than measured timing errors.

contrast-space overlap · hemodynamic response · sample-clock sensitivity · record removal · OpenNeuro ds000102