Audit date: 2026-08-22.
The data descriptor identifies Shared Reward as a block design: two runs, nine blocks per run, and eight trials per block. The public ds003745 events contain both block_<partner>_<predominant-feedback> rows and trial-level event_<partner>_<feedback> rows. Across the 100 tracked public run files inspected, event durations ranged from about 3.503 to 4.537 seconds (mean about 3.536 seconds). Run-level responded event counts vary because missed trials are represented separately.
The old convertSharedReward_BIDS.m explicitly used decision_onset plus trialDuration. That is evidence of the intended historical full-trial representation, not proof that exact outcome-phase timing can be reconstructed. Its questioned outcome-onset alternative and computer_non-faceclea typo are historical warnings, not instructions for a new source-data rewrite.
The current pooling candidate is an estimand-level full-trial model:
- ds003745 uses its published trial event rows unchanged;
- RF1 derives partner × feedback full-trial rows from validated canonical decision/outcome phase pairs;
- misses become one full-trial nuisance epoch only when both boundaries are present;
- source BIDS remains pristine.
Exact phase-resolved ds003745 events are neither required nor generated. Raw logs were not found in the checked repository or the public srndna-datapaper clone; their absence does not block the common full-trial model.
The old aging activation template has 13 EVs, 32 contrasts, 5-mm FEAT smoothing, and one unconvolved generic miss EV. Several named neutral/decision contrast vectors are malformed. The RF1 and r01-soi templates have 34 contrasts and nominal 6-mm smoothing. The pooled scientific definition must come from the authoritative crosswalk/model specification, not from the old aging FSF.
FLOBS variants are historical/sensitivity material. Old qsub, WarpKit, TEDANA, and fMRIPrep scripts document earlier processing but are not production inputs.
RF1 is multi-echo and uses TEDANA-enhanced confounds. ds003745 is single-echo and uses the corresponding fMRIPrep nuisance information. The current ds003745 wrapper pins the same fMRIPrep container version as Linux2 (25.2.5) and requires only MNI152NLin6Asym, but it does not pretend to reproduce RF1's multi-echo steps.
RF1 defines the common verified grid. ds003745 continuous BOLD is resampled by applying an identity transform with AFNI wsinc5 interpolation, and masks use nearest-neighbor interpolation, into a separate derivative. Grid equality is tested from dimensions, voxel sizes, orientation, and affine. Matched preprocessing uses classic combined FWHM as the 3dBlurToFWHM stopping criterion; ACF is retained as a diagnostic and for later residual-based inferential characterization. Phase 0 approved a 6-mm total classic-FWHM target on 2026-08-23 after all 865 characterization units passed. The target is supported by the measured distributions, voxel dimensions, and expected cross-age anatomical heterogeneity rather than inherited from historical FEAT settings.
- whether a block-level analysis should be a ds003745 sensitivity analysis only;
- final pooled covariates, including tSNR;
- pooled L3 design and subject exclusions;
- PPI harmonization and seed provenance.