Coupling and redundancy shape shock persistence in a computational model of social coordination
Journal of Computational Social Science
Anderson Díaz Pérez
Department of Social and Human Sciences, Center for Research in Life Sciences, Universidad Simón Bolívar, Barranquilla, Colombia
Correspondence: anderson.diaz@unisimon.edu.co

COMPUTATIONAL VERIFICATION

Full simulation: 43,200 rows, 42 columns, 120 base matrices and 600 matrix–coupling units.
Maximum absolute numeric difference from the archived run-level CSV: 0.0.
Maximum absolute difference in the normalized curves: 0.0.
Non-numeric CSV columns: identical.
This verifies a rerun of the supplied implementation, not an independent reimplementation or empirical validation.

JCSS ANALYSIS CORRECTION
The original analysis treated each matrix–coupling combination as a separate cluster and validation group. The JCSS analysis retains all coupling values of each base matrix in the same group.
The OLS point estimates are unchanged; clustered uncertainty and grouped predictive scores are recomputed. Family-wide and condition bootstrap intervals use base-matrix blocks. Family-by-rho intervals retain 20 independent seeds per family.
Grouped mean run-level R2: Family labels = 0.843296, Organization plane = 0.821485, Matrix descriptors = 0.816242, Shock + coupling intensity = 0.773755.
The weak-mode redundancy interaction has p = 0.1707 under corrected clustering.
Figure 5 envelopes are descriptive 10th–90th run-level percentiles. Figure 7 performance is calculated on matrix–coupling aggregates and differs from the mean run-level fold scores.

Observed environment: Python 3.12; numpy 2.3.5, pandas 2.2.3, scipy 1.17.0, matplotlib 3.10.8, statsmodels 0.14.6, scikit-learn 1.8.0, patsy 1.0.3.
