## ABSTRACT

The content of Section 2.5 and other observations about the impact of sampling on inferred network properties [112,154,162] makes a sampling theory for statistical network analysis one of the main priorities for developing the probabilistic foundations of the field; see Section 1.7.3 for further discussion. Over the next three chapters, I discuss network sampling and a number of related issues in some depth, culminating in a two‐stage formulation of a statistical model as a family of candidate distributions describing the uncertainty and variability in the data together with the context in which to interpret inferences made under the assumed model. These two aspects of modeling (uncertainty and context) come together in the concept of coherence. Incorporating the context and articulating the condition of coherence are two novelties of the framework presented below, cf. [52]. (See Chapter 5 for further discussion of uncertainty, context, and coherence. Refer back to Sections 1.4-1.5 for a high‐level overview of these essential modeling components.) To streamline the presentation, I tailor this framework to network models set either in a sampling context–called sampling models–or in a generative context–called generative models. I begin in this chapter with sampling models, and defer generative models to Chapter 4.