Measured trial amplitude means are missing from the examined ds000102 common estimate checks.

This publication examines whether a common fMRI estimate recovers half the sum of two realized condition means when trial amplitudes vary. It uses event designs from OpenNeuro ds000102 snapshot 00001 (52 runs, 1,248 trials) and synthetic signals; no brain images or measured brain amplitudes enter the calculation. Two candidate event fields disagree on 304 run-2 trials, including 296 correct trials. Under the specified model, row-choice weights give zero gain to each disputed correct trial. For synthetic amplitude changes bounded by ±0.1 units and constrained to sum to zero within each condition, the run-2 median maximum estimate change is 0.0946320932321282 units (maximum 0.10000000000000027), while both sample condition means remain fixed. Equal-gain weights recover the arithmetic mean across correct trials in all 52 designs. This mean equals the half-sum of condition means only in the 39 all-correct, balanced runs. The checks do not establish measured effects, population bias, biological causality, or likely synthetic changes. The dataset's exact OpenNeuro release is not identified beyond the recorded snapshot and Git commit.

functional magnetic resonance imaging · fMRI · trial amplitude · condition mean · event-related design · synthetic sensitivity analysis · estimand · label uncertainty · linear weights · OpenNeuro ds000102