At a 1.761 s lag, absolute correlations removed 21% of squared signed lag imbalance in five SEEG patients and 18% in eight fMRI participants

We measured the loss of signed lag imbalance when correlations become absolute correlations. Public data came from five SEEG patients and eight fMRI participants in separate movie studies. At a 1.761 s lag, an exact identity split squared signed lag imbalance into magnitude imbalance and an omitted sign term. Mean omitted fractions were 21.11% in SEEG and 17.52% in fMRI. Person bootstrap 95% ranges were 14.53-28.04% and 15.52-19.61%. All people exceeded the fixed 10% reporting target. The one planned joint test had p=0.03125. A broader three-test Holm check gave p=0.08203. The planned test of repeated omitted orientation did not pass in both modalities: secondary Holm p values were 0.75 and 0.05469. At descriptive lags of 5.283 s, mean loss reached 57.48% and 62.89%. An analytic Gaussian rotation had nonzero signed asymmetry but zero Gaussian mutual-information imbalance. Reversible finite records also showed sample sign loss. Thus, the result concerns information lost from an estimated score. It does not establish a shared neural source of time asymmetry, entropy production, or causality. Distinct cohorts, clinical sampling, filtering, motion, hemodynamics and short records limit interpretation. The package gives a diagnostic tool, fixed plan, exact inputs, source versions and terms, person results, negative checks, figures and repeat instructions.

human neuroscience · SEEG · fMRI · time reversal · lag correlation · sign loss · reproducibility