Analysis of Network Data

In this thesis we investigate the international arms trade network of major conventional weapons (MCW) between 1950 and 2013. After an introduction to the network theory and some descriptive analysis of the data we will model the arms trade network with the popular and well-known exponential random graph model (ERGM). However, we find that in order to guarantee a good model fit, the ERGM has to be extended into a curved ERGM. We are going to justify this step by introducing a method to model networks via generalized additive models (GAM), models which use smooth functions in order to include the effects of covariates. The estimated smooth functions will verify the use of geometrically weighted degree statistics in the ERGM. While discussing the use of GAMs for networks, we will also present a method based on a bootstrapping approach to model networks with logit models. Finally, we present and interpret the results of the fitted models.