Stored accuracy and feedback disagree on 625 money trial rows in ds005027 version 1.0.3

We checked all raw monetary incentive delay event files in OpenNeuro ds005027 and ds005012, both version 1.0.3. Stored accuracy and feedback disagree on 625 of 12,501 comparable money trial rows in ds005027. The conflicts occur in 63 files from 32 participants. No such conflict occurs in 9,600 money trial rows in ds005012. An example selection rule requires at least 50% success at each of two stored visits. Field choice changes inclusion for 5 of 72 ds005027 participants with complete scores and none of 60 ds005012 participants. No positive response time is available on any ds005027 row marked accurate. Feedback is an internal comparison field, not ground truth. The package preserves all 564 raw event files, 12 metadata files, fixed source hashes, audit code, full results, figures, and reproduction checks. A separate parser matches all 28,256 stored rows and 660 selection records. A clean rerun matches all ten numeric result files byte for byte.

monetary incentive delay · behavioral data · data audit · accuracy · feedback · response time · OpenNeuro