First run Stroop cost lowers observed test error in 69 selected TCP adults under fixed ridge models, but the approximate 95% gain interval includes zero

We tested whether first run Stroop cost helps predict second run cost after measured speed, accuracy, and background fields are known. The source was OpenNeuro ds005237 1.1.3. Fixed ridge models used 133 training people and 69 separate test people. Mean squared test error fell from 8681.49 to 6757.13 milliseconds squared, an observed decrease of 22.17%. The approximate 95% gain interval from 5000 refitted person bootstrap draws was -428.63 to 4750.75 milliseconds squared. This interval includes zero and leaves the added prediction gain uncertain. All six fixed checks gave positive observed gains, with intervals that included zero. Independent scoring and model checks passed. Two repeat analyses matched all 33 result files byte for byte. The package preserves 894 source files, trial decisions, person scores, code, predictions, bootstrap draws, provenance, and instructions. The result applies to selected adults within one session under the stated rules and independent person sampling assumption. It does not establish reliability across days, a brain mechanism, or a clinical marker.

Stroop task · cognitive control · response time · prediction · ridge regression · test–retest