Early rotation absolute error fell more than baseline error on day 2 in 32 ds004021 v1.0.1 people under a 40-trial rule

We tested whether early rotation error decreased more than baseline error on day 2 in OpenNeuro ds004021 v1.0.1. We used mean absolute error in fixed 40-trial windows. Valid initiation times were from 0.1 to 2.0 seconds. All 32 public people met the rule. The mean difference between error decreases was 6.86 degrees. The 95% paired person bootstrap interval was 4.41 to 9.31 degrees. Twenty-eight people had positive contrasts. All six planned checks gave positive contrasts with intervals above zero. First-trial error and initiation-time intervals included zero. These results do not establish an immediate benefit or unchanged timing. A check added after results kept all scheduled errors and bounded 12 missing errors from 0 to 180 degrees. It gave exact sample limits of 6.13 to 7.82 degrees. These are bounds for a different score, not confidence intervals. The source cohort already had published savings results. Our contribution is the baseline comparison, the checks, the missing-data proof, and the reusable package. The 32 people were selected from 40 recruits. The result does not identify a neural mechanism or a causal training effect. The package preserves sources, plans, code, all results, and excluded records. Fresh retrieval matched 132 source hashes. Complete runs in two Python environments matched all 68 result files byte for byte.

motor adaptation · visuomotor rotation · baseline adjustment · absolute error · initiation time · bootstrap · missing data · OpenNeuro