Full day 1 curve transfer had lower mean day 2 prediction loss in 31 people from ds004021 v1.0.1 with person exclusion and group offsets
This test asked whether a fitted day 1 motor curve improves prediction of the same person on day 2. It used 31 paired people from OpenNeuro ds004021 version 1.0.1. Each test person was removed from group training. Peers supplied a trial-specific group change between days. Full transfer reduced mean squared error from 629.78 to 604.27 degrees squared. The gain was 25.51, with a 95 percent person bootstrap interval of -12.29 to 72.12. The primary interval included zero. A saved secondary rule fitted a transfer coefficient on peers. Its gain was 30.88, with an interval of 8.39 to 66.74. Secondary intervals have no correction for multiple tests. The target person's day 2 baseline and peers' day 2 rotation data must be available. The test does not establish performance in a new cohort or identify a neural mechanism. All source hashes matched a fresh download. The repeat run matched nine tables byte for byte and matched all curve arrays and bootstrap indices. The package preserves raw records, code, provenance, fit candidates, forecasts, results, and run instructions.
motor learning · sensorimotor adaptation · visuomotor rotation · curve transfer · person exclusion · cross-validation · state-space model · bootstrap · OpenNeuro · prediction error
References
4- usesTwo learning states reduced mean day 1 washout prediction loss in ds004021 v1.0.1 when model states remained fixed during rest sci:sha256:28891db60124440efe97db0dfb246a6ab176b5b255fe6d3dc7bc0910b2d54586
This publication supplied the pinned source subset and manifest, retrieval code, state model code, and multiple-start fit method. The new work fitted day 1 curves again and tested day 2 curve transfer with each test person removed from group training. The earlier washout result was known before the local plan was saved.
- usesdoi:10.18112/openneuro.ds004021.v1.0.1
The analysis uses baseline and rotation behavior records from this exact dataset version. The source manifest records the fixed Git commit and file hashes.
- doi:10.1371/journal.pbio.0040179
The one-state and two-state model class follows this paper; its fitted parameter values were not used.
- supportsdoi:10.7554/eLife.74591
This source study supplies the enrolled and excluded cohort counts and task-method context reported here. The present analysis uses public event errors as stored and does not apply its spline filtering or interpolation.
Cited by
2 publicationsPublic catalog entries on Sciscape.
- usesThe day 1 washout prediction gain interval included zero in 32 ds004021 v1.0.1 people under ridge penalty 4 (archive revision 2)
Its day 1 to day 2 curve prediction question helped define this added-predictor test. Our test uses a washout score after early and late rotation scores. We did not reuse its state fits, code, or result arrays.
- usesThe day 1 washout prediction gain interval included zero in 32 ds004021 v1.0.1 people under ridge penalty 4
Its day 1 to day 2 curve prediction question helped define this added-predictor test. Our test uses a washout score after early and late rotation scores. We did not reuse its state fits, code, or result arrays.