An equal personal and peer score average reduced later flanker error with 20 trials per condition in 133 selected ds004883 1.0.1 adults

This test used the first 20 valid early trials per condition to predict later flanker conflict costs. It used OpenNeuro ds004883 version 1.0.1 under stated event and selection rules. The selected group contained 133 adults and three tasks. An equal average of personal and excluded-person peer scores reduced mean squared error from 1416.5 to 630.9 ms squared. The reduction was 55.5%, with an approximate 95% refit person-bootstrap interval of 49.5% to 60.4%. All six planned checks favored this rule over the raw score. The main rule did not clearly outperform the peer mean alone. Algebra and all task pair comparisons showed unchanged score orders and correlations. Thus, lower error did not show improved rank reliability. This is a within-task, within-visit test in one selected public sample. It does not identify a latent trait or establish use in a new cohort. The package preserves source bytes, a fixed local plan, code, trial selections, bootstrap draws, figures, and reproduction instructions. Fresh downloads matched all 1041 source hashes. A full rerun reproduced all 41 analysis files byte for byte.

flanker task · response time · prediction error · peer average · within-task prediction · cognitive task scores · measurement noise · OpenNeuro