This chapter addresses model specifications for the co-evolution of a social network and a discrete ordinal (or dichotomous) behavior variable. As before, this is done by means of an evaluation function composed of effects. Effects here are functions of the behavior, possibly also the network, as seen from the focal actor. Change statistics here are defined as changes in the value of the effect when the behavior variable is increased by one unit. The chapter defines several effects: first, depending only on the behavior of the focal actor; subsequently, depending on the behavior as well as the network. The latter kind of effect represents social influence or generalized social influence. Before such effects are proposed, it is mentioned that three basic aspects of social influence must be chosen: the reference group, the mechanism of social influence in a dyad, and aggregation over multiple dyads in the reference group. For the mechanism of social influence, a distinction is made between distance reduction (assimilation) and correlational influence. Effects are mentioned for both.

The stochastic actor-oriented model is a statistical model for network dynamics, meant to be used for network panel data, that is, repeated measurements of a network on a given set of actors. This chapter gives an introduction to the probability model, in three flavors: a short intuitive explanation, a long conceptual one, and a mathematical explanation. These alternative, complementary formulations cater to readers with different interests, backgrounds, and preferences. The model assumes that between the moments of measurement, the network changes gradually in unobserved small steps. The heart of the model is the ministep, playing a basic role together with the evaluation function and the rate function. The ministep is the fundamental event in which actors have the opportunity of changing one network tie; the subsequently measured networks are connected by a sequence of unobserved ministeps. The probabilities of network change in the ministep depend on the evaluation function according to a conditional multinomial logit model. The rate at which ministeps take place is given by the rate function. The model is also formulated as a simulation model. The chapter ends with a brief history of the model.