Testing the Dependence between Network Positions and National Attributes in Countries Interaction Network

Exploring the relationship between the positions of countries in the countries interaction network and their development status is of great significance in world system studies. To characterize the complex interactions between countries, network methodologies have been extensively employed. The methods typically used in the field include blockmodel analysis and statistical models, which have the deficiencies of instability, dependency on specific assumptions, and requirement of domain knowledge. Enlightened by the network embedding methods, we propose a novel framework combining network embedding and correlation analysis for addressing the problem in a model-free data-driven way, and provide a quantitative test statistic representing the magnitude of the correlation. We test our method on a real world dataset containing countries’ trade and political relations as well as several indices of development levels. Experiment results derived from our method are in consistency with previous findings via other approaches, and provide extra information about the pattern of the relationship.