The 95 percent interval included zero for the gain from flanker time variation in ADHD symptom prediction in 70 ds005754 v1.1.1 test youths with fixed models and independent person sampling

We tested whether flanker response time variation added ADHD symptom prediction information in OpenNeuro ds005754 version 1.1.1. The target was the raw sum of 18 supplied parent items. The baseline linear model used age, sex, speed, and conflict. The full model added log time variation within conditions. A fixed internal split used 104 training youths and 70 test youths. Test mean squared error fell from 162.785 to 159.782 score points squared. The gain was 3.003, or 1.845 percent. The approximate 95 percent test-person bootstrap interval was -1.630 to 7.581. It held models fixed and assumed independent, comparable people. The training and test refit stability interval was -17.959 to 13.918. All six fixed check intervals included zero. A secondary whole-cohort omitted-person test gave a gain of 0.051 score points squared. The evidence does not establish a useful gain or the absence of a gain. This is an extension in a previously studied cohort. The local plan, all 197 source files, code, provenance, trial decisions, predictions, bootstrap draws, and reproduction commands are preserved. Separate calculations passed. A fresh public repeat matched all 26 analysis files byte for byte.

response time variation · flanker task · ADHD symptoms · prediction · bootstrap · youth