Nearly any entity in the modern world can be seen as a component of some network, ranging from transportation, telecommunication, and brain networks to crime location and socio-economic networks. Given recent technological advances in data collection and computing, networks as data structures are increasingly more widespread as well. However, while the number of empirical applications employing networks is rapidly increasing, the underlying theoretical intricacies and implied concerns remain less well-understood. It is important to recognize how much more complex and nonlinear networks are when compared to classical data structures in econometrics.
In his current research, Julius Vainora is working on several crucial econometric aspects of network data with emphasis on dependence and statistical inference. As an initial step, he builds a formal methodological framework to handle statistical relationships between observed characteristics of network entities and explores its theoretical properties. Using graph theory along with this formal language, a number of asymptotic results essential for statistical inference in applied work are derived. It demonstrates that credibility of empirical results in the context of networks depends dramatically on the nature of the network and the underlying dependence structure. The research is complemented with various empirical tools, their implementations, and real data applications.