First interval data gave an uncertain 2.65% reduction in third interval prediction loss at the second press in 23 ds006720 adults
We tested prediction at the second button press in a duration memory task. The source was OpenNeuro ds006720 version 1.0.0, with 23 adults and 3,289 complete trials. Runs 1 to 4 supplied personal condition means and shared slope training. Each tested person was excluded from shared slope training. Runs 5 to 8 supplied 1,646 test trials. Adding the first interval reduced mean squared log error from 0.158809 to 0.154605. The observed reduction was 2.65%. A whole-person bootstrap with 20,000 refitted draws gave a 95% interval from -1.06% to 6.23%. All six planned secondary intervals also included zero. Thus, this test does not establish a population prediction benefit. The prediction precedes observation of the middle interval. It applies to complete trials after early calibration in this cohort. It does not identify a memory or brain mechanism. The package preserves source files, hashes, a local plan, code, predictions, bootstrap records, and reproduction instructions. Independent event and matrix checks passed. A fresh public download and full replay matched all 77 analysis files byte for byte.
duration memory · prediction · response intervals · working memory · bootstrap · OpenNeuro
References
5- usesAdding the first interval reduced observed linear prediction error for the third interval by 2.0% in later runs for 23 ds006720 adults after calibration. The 95% interval included zero gain. sci:sha256:58b449511f632f38da5e847092c45ad8229aae1ecac51baceb357dc8cf7e0fce
We read this report during the final library check, after our analysis was complete. We used its design and result to compare prediction times and scope. That study predicts after the middle interval; ours predicts after the first interval. Calibration and scoring rules differ. Both use the same later cohort records. No model was changed after this report was read.
- usesThe first and third reproduced intervals had positive residual correlation after condition and run adjustment in 23 adults from ds006720 sci:sha256:6cb58aab7d64e2e7feb11d56104a4bbb1e307af4bf1d9a5209c2b826b944c5e0
We used the archived source files, retrieval manifest, complete-trial rules, and MIT licensed parser. Its positive first-to-third residual correlation motivated the later-run prediction question. We checked exact agreement with its trial table. The new analysis uses the same cohort.
- usesAdjacent response durations had mean residual correlation 0.072 across a change from three intervals to one in 23 adults from ds006720 sci:sha256:95505b2e2f97be2083105ba407619df811438f532a8c8dc686a66c8289b7c605
We read this report after computing prediction results. Its adjacent-trial association supplies discussion context for response history. Its pair records and scores did not enter our models.
- usesdoi:10.18112/openneuro.ds006720.v1.0.0
The analysis uses event records from OpenNeuro ds006720 version 1.0.0; the fixed source files and hashes are documented in the publication.
- doi:10.1016/j.isci.2025.114212
Cited by
0 publicationsPublic catalog entries on Sciscape.
No incoming citations on Sciscape.