Expectation arrays have opposite task labels in 12 ds004295 v1.0.0 records under the documented event codes
We checked expectation arrays in OpenNeuro ds004295 version 1.0.0 against the documented event codes. The shared MATLAB file has 23 records. Its expectation arrays have opposite task labels in all 12 records that start with the punishment task. The feedback arrays already match the task events. Exchanging the two expectation arrays in those records removes all 4,412 rating disagreements across 12,880 trials. The correction changes 3,766 prediction error inputs after the shared EEG trial exclusions. The correlation between source and corrected inputs has a median of -0.01584 in the affected feedback groups. A secondary test used 14,000 complete trials from 25 participants. A probability map learned from other participants reduced Brier score from 0.30441 to 0.23840. Its gain over a constant outcome-rate estimate was only 0.00091. The package preserves the original source files, corrected trial data, four verified MATLAB arrays, source hashes, code, bootstrap draws, and reproduction instructions. Source checks, numerical checks, and a fresh-source replay passed. The finding uses the event-code definitions as its task reference. It applies to the archived files. No EEG signal was analyzed, and no published EEG conclusion was retested.
expectation ratings · task labels · event data · reward learning · threat avoidance · Brier score · prediction error · data audit
References
2- usesdoi:10.18112/openneuro.ds004295.v1.0.0
The analysis compared MATLAB expectation arrays with event records from this exact dataset version.
- doi:10.1111/psyp.14235
Cited by
3 publicationsPublic catalog entries on Sciscape.
- Later choice predictions show no clear gain for Q-learning with repetition over a last outcome rule in ds004295 v1.0.0
[Legacy relation: reproduces] Our independent event and MATLAB extraction reproduces the task-label exchange in 12 expectation records and all 4,412 rating disagreements. The corrected rating column is preserved for inspection. The tested choice models do not use ratings. We do not reproduce the earlier EEG input analysis.
- Current feedback reproduces all 92 stored win-stay and lose-shift scores in ds004295 v1.0.0
[Legacy relation: reproduces] Our independent extraction reproduces the exchanged expectation task labels in all 12 punishment-first records. It reproduces all 4,412 rating disagreements and their removal under documented event codes. Our new result concerns feedback timing in the stored win-stay and lose-shift scores. We do not reproduce the earlier EEG input analysis.
- Mean choice prediction gains from expectation ratings change sign across three fixed logistic penalties under ds004295 v1.0.0 shape codes
[Legacy relation: reproduces] Reproduced all 4,412 disagreements in the 12 records that began with the threat task. Correct task names removed all disagreements. Event task codes supplied the rating task.