A Nash Three-Player Game on Null Hypothesis Decisions and Perturbed Garwood Bounds for Big Health Data
IIn large-scale health datasets, classical frequentist hypothesis testing against a zero-effect null routinely yields very small p-values as sample or population size grows, even when the observed event count is modest and structurally fragile. While a recently proposed Garwood-based structural index evaluates signal resilience independently of population size, and a perturbation parameter measures how much the Garwood lower bound must shift to match the frequentist verdict, comparing these perspectives requires a formal common ground. We model the interaction among the classical null test, the perturbed Garwood index, and the bound perturbation we have proposed as a three-player game. Using Kakutani's fixed-point theorem on bounded, convex strategy spaces, we prove that a Nash equilibrium always exists, turning the equilibrium perturbation and its leverage into a diagnostic measure of the bound adjustment needed to reconcile the structural index with the frequentist decision to reject a null hypothesis until high event counts reach the saturation threshold where the game ceases to be informative.
Authors
- Marco Roccetti (ORCID: https://orcid.org/0000-0003-1264-8595)
Institutions
- University of Bologna (IT)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-21
- DOI
- https://doi.org/10.5281/zenodo.24425648
- Primary Topic
- Probability and Statistical Research
- Type
- preprint