Twenty changes from trial_type made the primary run correlation negative in the ds000102 snapshot 00001 label model

For 26 participants in OpenfMRI/OpenNeuro dataset ds000102 snapshot 00001, we calculated the minimum balanced condition-label changes that could make the Pearson correlation between run-specific raw reaction-time scores nonpositive under a fixed label model. With correct, finite, positive RTs, the minimum was 20 changes from `trial_type` and 26 from `Stimulus`; adding an RT ≤ 2000 ms limit changed these minima to 14 and 6. The thresholds are conditional on retaining run-1 labels, agreeing run-2 labels, complete-run condition balance, and at least eight usable responses per condition. Changes are not measured label errors, and the selected assignments do not identify correct labels. Exact rational optimization found a positive minimum covariance at each preceding feasible cost. The result is limited to this sample and model; it does not estimate population reliability or biological causality.

reaction time · condition labels · Pearson correlation · sensitivity analysis · exact optimization · Flanker task · individual differences · ds000102