Random effects in dynamic network actor models

Abstract Dynamic Network Actor Models (DyNAMs) assume that an observed sequence of relational events is the outcome actor-oriented decision process consisting two levels. The first level represents time until actor initiates next event, modeled by exponential distribution with actor-specific activity rate. second describes choice receiver a conditional multinomial logit model. DyNAM assumes parameters are constant over actors and context. This homogeneity assumption, albeit statistically computationally convenient, difficult to justify, e.g., in presence unobserved differences between or contexts. In this paper, we extend DyNAMs including random-effects vary across contexts allow controlling for unknown sources heterogeneity. We illustrate model analyzing among users online community aspiring professional digital graphic designers.