Replication Package: Longitudinal Clustering of National Innovation Systems (GII 2016-2025)

Description

Technical info (English)

This repository contains the complete R code and dataset to fully reproduce the results of the study: "From Polarization to Rigid Equilibrium: A Robust IT-Driven Clustering of National Innovation Systems".

The study develops a robust framework for longitudinal clustering of panel data, integrating Hopkins statistic, consensus clustering (NbClust), Ward's method, and temporal stability metrics (ARI, Baker's Gamma). The empirical analysis covers 100 countries over 2016-2025 using the seven pillars of the Global Innovation Index.

Files included:
- Full_etap.R: Complete R script implementing Stages 1-4 of the methodology
- ALL_2.csv: Panel dataset (GII 2016-2025, 100 countries, 7 pillars)
- /table: Generated tables (Hopkins statistics, NbClust results, ARI dynamics)
- /figure: All visualizations (8 figures, including alluvial plot, heatmap, radar charts)

Key findings reveal the evolution of the global innovation landscape from binary polarization to a rigid three-cluster model (Leaders, Followers, Modest Innovators), with the COVID-19 pandemic acting as a structural shock. The case study of Ukraine demonstrates the "Efficiency Anomaly" — high innovation outputs despite critically low institutional quality.

All results are fully reproducible. The script automatically installs required R packages and generates all tables and figures.