Pre-Referendum Solar Wind Density Variability and Geomagnetic Deviation Signals_Dataset

Two space weather signals replicated across independent time windows, held under restriction to a single political event subtype while failing in two parallel subtypes tested under identical conditions, and reappeared in a subsequent broader population validation under stricter methodological checks. That convergence is the primary finding of this study. The signals were identified in a corrected Wikidata event inventory spanning December 2009 to May 2026, with replication windows of 2010 to 2017 and 2018 to 2026. A subtype audit retested proven politics signals within referendum, treaty, and coup event subtypes at identical pre-specified leads across both windows. Solar wind proton density variability, measured as Density_var6h_max, replicated at T-14 in both windows with AUC values of .549 and .567 and false discovery rate values of .024 and less than .001. A geomagnetic deviation measure, self_dual_distance, replicated at T-9 in both windows with AUC values of .559 and .536 and false discovery rate values of .004 and .088. Neither signal replicated across both windows in treaty or coup subtests. A subsequent validation on the broader all-politics population recovered both signals as clean survivors after parity, support, non-latency, and autocorrelation checks; self_dual_distance retained T-9 lead consistency, while Density_var6h_max persisted in the same direction but shifted lead in the broader population validation. Effect sizes are small and fall below a prespecified practical effect floor of AUC .580. The subtype concentration and cross-configuration persistence, not the effect magnitude, are what warrant reporting. The findings support a narrow association-level claim and do not imply causality or operational forecasting skill. Keywords: referendum, space weather, solar wind density, geomagnetic activity, Wikidata, false discovery rate, cross-window replication