OLS scale gave 7.05% null rejection in a fixed AOMIC contrast test with simulated Gaussian damped noise (q=0.9, lag-one covariance=0.4)
We tested 95% contrast intervals on 24 fixed AOMIC designs under five specified Gaussian noise rules. We used the earlier weights with a bound on contrast variance. An OLS residual scale estimate gave excess null rejection in six cases. Simultaneous 95% lower bounds exceeded 5% in these cases. The largest observed rate was 7.05% in 20,000 null draws. The pooled rate was 4.62%, which hid these failures. We also give an unknown-scale interval and a coverage proof under valid covariance bounds. This interval gave no rejections in 2.4 million null draws. Its mean radius was 9.03 times the correct-covariance reference radius. It had zero observed power at the tested effect size. No BOLD image values were used. Code, fixed designs, source hashes, seeds, results, and repeat checks are included.
fMRI · contrast intervals · temporal covariance · residual scale · Gaussian simulation · coverage · AOMIC
References
1- usesNew weights gave lower maximum contrast variance than AR(1) in 24 fixed AOMIC designs over damped cosine mixtures with known variance=1, lag-one covariance=0.4, and q=0.9 sci:sha256:dbf9bc4d0cd1c25a3afbd309fa14ae5e9330e40fc484d6fb143af59a802d5f56
Uses its 24 fixed AOMIC designs, source archive, q=0.9 protected weights, variance bounds, and stationary covariance oracle. Tests an unknown residual-scale inference rule that the parent did not claim to calibrate.
Cited by
0 publicationsPublic catalog entries on Sciscape.
No incoming citations on Sciscape.