A Pre-Registered Framework for Composite Sensitivity Analysis Under Explicit Null Models

This work presents a pre-registered composite sensitivity framework for exploratory statistical analysis of cross-domain correspondences under explicitly defined null models. The framework formalises hypothesis specification, null model construction, and composite scoring procedures in a reproducible structure designed to minimise post-hoc flexibility. A case study is included to demonstrate application under nine pre-registered hypotheses derived from a historical text and evaluated against publicly available astrophysical and geophysical datasets. All results are conditional on the adopted modelling assumptions, mapping rules, and null model specifications. The study does not claim causal, predictive, or physical interpretation of the observed correspondences. This deposit includes the final PDF version of the manuscript and is released for transparency and reproducibility purposes. Source code and repository:https://github.com/sidichaker1-beep/chigar2026_sensitivity