台大心理系

A new item response theory model was developed in this study to model social networks. The model can handle both directed and undirected networks. For directed networks, the model assumes that a valued edge is determined by the sender's latent and observed preferences and the receiver's latent and observed features. For undirected networks, the model assumes that actors' latent preferences and latent features are equal. The model was built under a Bayesian framework, and Markov chain Monte Carlo methods were used to approximate the full conditional posterior distributions. The simulation study showed that the model parameters were satisfactorily recovered. The applicability of the model has been demonstrated by implementing it with empirical data.