Mixed logit models and network formation

Abstract The study of network formation is pervasive in economics, sociology, and many other fields. In this article, we model as a ‘choice’ that made by nodes to connect nodes. We these ‘choices’ using discrete-choice models, which agents choose between two or more discrete alternatives. employ the ‘repeated-choice’ (RC) formation. argue RC overcomes important limitations multinomial logit (MNL) model, gives one framework for studying formation, it well-suited also illustrate how use accurately both synthetic real-world networks. Using edge-independent networks, compare performance MNL model. find estimates data-generation process our networks than patent citation network, forms sequentially, present case qualitatively interesting scenario—the fact new patents are likely cite older, cited, similar patents—for employing yields insights.