Fixed donor weights raised mean error estimates at 24 selected sites in six SEEG patients and 16 fMRI adults

We tested whether donor repeat disagreement can set useful repetition weights for other people. Separate public studies supplied human SEEG high-gamma amplitudes and BOLD fMRI estimates. Two patients and four MRI adults supplied donors and development. A local code lock preceded scores in six other patients and 16 other adults. Each person supplied 24 source-selected sites and 165 sounds. The fixed rule used square-root donor weights with 50% shrinkage. Each sound group kept its total budget. Mean estimated gains against uniform weights were -0.65346% and -0.19524%. Their pointwise 95% person Student intervals were [-0.89556,-0.41136]% and [-0.28689,-0.10358]%. Negative gain means higher estimated error. A shared negative-mean test passed after three-claim Holm correction, p=0.046875. This direction was added after donor checks, before reserved scores. The positive-gain test failed. The fixed mean-below-1% test passed. All negative outcomes and controls remain. Unchanged MRI weights gave mean gain -0.31293% in eight records from another study. An exact identity separates variance alignment from the cost of unequal allocation. Even correlation one can give a worse budget. This is a retrospective weighted-disagreement benchmark. Future error needs independent, stable repeats and fixed processing. Source screens, shared electrical orders, separate cohorts, sparse coverage, short baselines and unequal measurement errors limit that model. The result does not establish optimal budgets, actual recording-time savings or a common neural cause. Code, aggregate profiles, source pins and terms, all results, figures and replay steps are supplied. Original neural arrays stay external.

human neuroscience · SEEG · fMRI · repeat disagreement · repetition allocation · measurement precision · independent cohorts · negative result