In ds000102 snapshot 00001, run 2 has a larger mean CV than run 1 after 13 selected trial removals.

We tested whether the mean run-2 minus run-1 pooled coefficient of variation (CV) in 26 participants could change sign when at most one retained trial was removed per participant. In the primary filter (correct responses, 200–2,000 ms), the mean CV change was −0.021640247 before removal. Its maximum was −0.000207536 with up to 12 participant removals and +0.000879802 with up to 13. Thus, 13 is the minimum number of removals that can make this mean positive under the specified rule. All 13 maximizing removals came from run 1. In a sensitivity filter of all correct responses above zero, the minimum was 10 removals. These are selected, adversarial removals; they do not identify incorrect trials. The result applies to this sample, filters, pooled CV definition, and removal rule. It does not establish a biological cause or a population effect. The analysis used OpenNeuro ds000102 snapshot 00001, pinned to a recorded Git commit; the exact OpenNeuro release associated with that commit is not established.

coefficient of variation · response time · trial removal · sensitivity analysis · Flanker task · OpenNeuro · robustness · Monte Carlo