Task block selection added 232 eligible people under fixed scoring rules in Healthy Brain Network contrast task releases 1 to 4, version 1.0.1

This study used Healthy Brain Network contrast task records from releases 1 to 4, version 1.0.1. The original whole file rules selected 244 people with three valid runs and at least 10 correct run 1 response times. An extra method selected the unique task block named in each file. It added 232 eligible people, for a total of 476. All 17,568 original target records stayed unchanged. The audit found 272 repeated person and block groups with identical relative task marker sequences. This finding concerns event records and does not establish repeated EEG signals. The planned forecast trained on releases 1 and 2 and tested on releases 3 and 4. Adding response time spread reduced mean log loss by 0.002646 nats per target. The 95% bootstrap interval was -0.003290 to 0.008506. All eight planned checks also had intervals that included zero. The extra sample gave a gain of 0.001289, with an interval of -0.001843 to 0.004552. Thus, neither sample resolved a forecast gain from this input. The extra method was added after the main result was known. The package preserves raw source records, fixed versions, hashes, scoring code, exclusions, models, and bootstrap draws. Fresh analyses reproduced all numerical results. The study does not test neural causes or clinical diagnosis.

Healthy Brain Network · contrast change detection · behavioral prediction · response time spread · task block selection · event data · ridge logistic regression