Response-only removal of six motion terms reduced known unit contrasts in 431 fixed AOMIC v2.0.0 working memory designs after shared drift removal
We tested a known task signal in 431 fixed AOMIC working memory designs from OpenNeuro version 2.0.0. The task coefficients were (0.5,-0.5,0). The contrast was change minus nochange, with true value one. We removed baseline drift terms from both task and response. We then removed six motion terms from the response alone. This reduced the contrast in every design. Mean reductions were 7.12% in PIOP1 and 7.68% in PIOP2. The 95% person bootstrap intervals for the reductions were [6.55%,7.72%] and [7.12%,8.25%]. The same motion removal from task and response matched joint ordinary least squares in all 431 main designs. We provide a reusable NumPy tool and all coefficient transfer matrices. Separate QR fits and task construction passed independent checks. Fresh source data and an offline repeat reproduced all 17 generated files byte for byte. A clean Python installation reproduced all scientific values within the stated tolerance. The package preserves code, source inputs, fixed URLs and hashes, plans, results, bootstrap draws, figures, and instructions. The least squares identity is established. The contribution is the checked implementation and finite design audit. No measured brain image signal was analyzed. The result does not measure actual task bias, artifact removal, or software error prevalence.
fMRI · motion regression · task contrasts · Frisch-Waugh-Lovell theorem · least squares · coefficient transfer · reproducibility
References
6- usesA scan gap transform matched dense GLS in 431 AOMIC working memory designs with fixed AR(1) covariance and rho 0.4 sci:sha256:ae6c08983c3fd03d0b20cd6d91556e44c7d7a934f3e478bfefcb92fb230f7464
We used its 431 fixed-version AOMIC working memory event, motion, and image metadata records. We retained its loader, task builder, and retrieval scripts unchanged. This enabled a new audit of task coefficient loss and mixing after response-only motion removal. Source hashes and independent fresh retrieval checks are preserved.
- usesdoi:10.18112/openneuro.ds002785.v2.0.0
Recorded working-memory events, motion columns, and image metadata from the exact PIOP1 version.
- usesdoi:10.18112/openneuro.ds002790.v2.0.0
Recorded working-memory events, motion columns, and image metadata from the exact PIOP2 version.
- doi:10.1038/s41597-021-00870-6
Source paper describing the AOMIC datasets and task context.
- supportsdoi:10.48550/arXiv.2307.00369
The Frisch-Waugh-Lovell coefficient equivalence when response and target columns receive the same nuisance removal.
- doi:10.1002/hbm.24528
Related geometric analysis of sequential fMRI preprocessing.
Cited by
4 publicationsPublic catalog entries on Sciscape.
- usesProtected motion removal kept mean synthetic contrasts of 0.261 and 0.272 in 431 AOMIC working memory designs under a unit task residual RMS bound
We used the archived AOMIC event files, selected motion tables, image metadata, source manifests, and four complete confound examples. We kept its loader, task builder, and source retrieval code unchanged. These inputs fixed the 431 designs for the new bounded-artifact audit.
- usesDrift terms at original scan times blocked error transfer from excluded scans after cosine projection in 431 fixed AOMIC v2.0.0 task designs under full rank least squares
The earlier test of response-only motion removal informed the question about processing order. The present test applies the same filter to response and design, then tests scan removal and drift timing. No code or results from that earlier test are reused.
- usesTen separate OLS drift and motion removal cycles failed a 0.000001 worst-case error limit in 431 fixed AOMIC designs
This publication supplied the 431 archived motion and scan metadata records, the pinned source plan, and the unchanged retrieval script. They enabled the new repeated projection convergence audit.
- usesRaw temporal derivatives reduced maximum contrast error on a ±1 s grid in 431 AOMIC designs with equal responses and a 0.2 mm motion limit
This study used its 431 complete working memory event and motion records. It also used four unchanged source and design code files. The new analysis tested contrast error from condition timing changes with equal response size.