Linear response filling reduced a known contrast on average in 431 AOMIC working memory designs with unchanged full designs and a 0.2 mm motion limit

We tested linear filling of removed response samples in 431 public AOMIC working memory designs. The full fit designs stayed unchanged. The primary rule removed the first scan and scans with framewise displacement above 0.2 mm. All response signals were constructed; no BOLD image was used. The mean loss of a known unit task contrast was 5.10% in PIOP1 and 2.11% in PIOP2. Of the 431 runs, 374 had a loss, 23 had a gain, and 34 were unchanged within the stated tolerance. We calculated sharp false-contrast bounds for modeled nuisance signals with a true task contrast of zero. The mean bound was 0.35656 and 0.25005 when nuisance RMS on retained scans was at most one. Saved signals reached these bounds. Filling only the task columns preserved the known task contrast but left a nonzero nuisance bound. An added check with an RMS limit over the full run had mean bounds of 0.26063 and 0.20721. Independent checks passed for 3,447 valid run-setting fits and 431 primary bound signals. Fresh-source and offline repeats matched all 11 result files byte for byte. The package contains source inputs, hashes, code, proofs, results, figures, and repeat instructions. The findings apply to the stated designs, masks, model, and fill rule. They do not estimate a neural effect in measured fMRI.

functional MRI · motion censoring · linear interpolation · task contrast · nuisance signal · signal processing · AOMIC