ABSTRACT
The choice to use a generative model in place of a sampling model reflects the perspective from which the network is being analyzed. Generative models treat the observed network as evolving, while sampling models treat the observed network as having been sampled from a population network. This chapter discusses the automatic consistency properties of generative models, both in terms of generative consistency and consistency under selection. Because of this connection, the language of generative models is often invoked when modeling real-world networks, even in the absence of any natural interpretation for the generating mechanism in the given context. Although generative models are of interest for predictive modeling and machine learning applications, the chapter emphases on sampling models reflecting the intention to focus on those essential elements of network analysis which have not been given due attention elsewhere.
