Statistical Network Models

Abstract

As suggested in the previous chapter, for most of the history of network science analysts were limited to network visualization and network description. Only in the past couple of decades has network statistical theory and computational power developed enough to allow for valid and feasible statistical modeling of networks. The primary barrier to statistical network modeling was the fundamental assumption of independence of observations that underlies much of traditional statistical theory. Networks by definition are non-independent. If you know that one actor is tied to another actor, you have information about the second actor that is dependent on the first.

Prediction and explanation are exactly symmetrical. Explanations are, in effect, predictions about what has happened; predictions are explanations about what’s going to happen. (John Rogers Searle)