{"country":{"country_code":"RW","incidents_365d":6,"leads":[{"corr":0.6748,"country":"AZ","lag_days":10,"p_adj":0.0},{"corr":0.6477,"country":"BY","lag_days":5,"p_adj":0.0},{"corr":0.6474,"country":"TZ","lag_days":1,"p_adj":0.0},{"corr":0.6074,"country":"JO","lag_days":8,"p_adj":0.0},{"corr":0.5707,"country":"EG","lag_days":10,"p_adj":0.0}],"led_by":[{"corr":0.5439,"country":"SY","lag_days":19,"p_adj":0.0},{"corr":0.5241,"country":"PK","lag_days":30,"p_adj":0.0},{"corr":0.4895,"country":"MM","lag_days":30,"p_adj":0.0},{"corr":0.4354,"country":"IQ","lag_days":14,"p_adj":0.0}]},"generated_at":"2026-05-21T12:55:40.096762+00:00","honest_caveat":"Cross-correlation does NOT imply causation. Significant lead/lag pairs may reflect (a) one country influencing another, (b) shared regional or geopolitical drivers (elections, sanctions, fiber routes), or (c) downstream measurement effects from overlapping ISP infrastructure. lag=0 (simultaneous) is intentionally excluded -- see /atlas/correlation-matrix for that case. With 50 countries and 31 lags evaluated per pair, the raw p-values overstate significance; we apply Benjamini-Hochberg FDR control at alpha=0.05 and require |r| >= 0.4. Counts come from the citable-censorship subset (incident_type IN censorship/mixed) -- IODA disruption rows are excluded since they include fiber cuts / weather / DDoS.","method":"pearson-cross-correlation-bh-fdr","version":"v1"}