Original scan gaps reduced null rejections with grid REML in 431 fixed AOMIC designs under Gaussian AR(1) noise at rho 0.6 independent of masks

We tested whether fitted noise correlation compensates for lost scan gaps after motion scan removal. We used 431 fixed AOMIC working memory designs and fixed motion masks. Gaussian AR(1) noise was independent of the masks. The primary noise coefficient was 0.6 per original scan. Each design had 1,000 null responses. Both fitted methods used the same restricted maximum likelihood grid and nominal two-sided 5% t rule. Mean rejection rates were 6.447% with consecutive retained rows and 5.723% with original scan gaps. The paired difference was 0.724 percentage points. Its conditional 95% Monte Carlo interval was 0.686 to 0.762 percentage points. The known-covariance reference rate was 5.044%. Keeping scan gaps reduced error, but the fitted test still exceeded 5%. Four secondary settings and an exact repeat support the benchmark. The package contains a checked NumPy tool, fixed sources, all results, and reproduction instructions. The result applies to the stated finite grid and simulated noise.

AR(1) noise · fMRI · motion censoring · restricted maximum likelihood · scan gaps · null rejection rate