Removal of one response decreases the pooled CV by 43.36% in ds000102 sub-01 run 2

This analysis examines how removal of one recorded response changes pooled response-time scores and participant ranks in 26 participants performing two runs of a flanker task. The data are 56 event and documentation files from OpenNeuro ds000102 snapshot 00001, identified by the recorded Git commit and source hashes; the exact mapping from this historical snapshot identifier to a versioned dataset DOI is uncertain. No brain images or demographic files were used. The primary filter includes 1,230 correct, finite, positive response times. A sensitivity filter includes 1,228 correct responses from 200 through 2000 ms. For each participant and run, each retained response was removed in turn, with a no-removal case. In sub-01 run 2, removing one correct 1,022 ms response reduced pooled coefficient of variation (sample SD divided by arithmetic mean) from 0.27653850931873974 to 0.1566321595350499, a 43.359729565000715% reduction; the mean decreased by 4.903206097051163%. Across participant pairs, the primary analysis allowed opposite CV order for 120/325 run-1 pairs and 77/325 run-2 pairs. Under the sensitivity filter, the counts were 121/325 and 78/325. These pair-specific possibilities do not describe one shared removal case. For sub-20 run 1, the primary CV rank was 19 and its possible range was 2 to 24. Arithmetic, pair, and rank outputs passed the saved checks, which were performed by one researcher and are not independent verification. These finite-sample ranges are not population confidence intervals. The findings do not establish that influential responses are incorrect, resolve condition labels, or show brain–behavior relations, biological cause, or separate-session reliability. The same 26 participants were used in earlier reports; this analysis adds no biological sample.

response time · coefficient of variation · participant ranks · response removal · flanker task · OpenNeuro · finite-sample sensitivity