Dynamic Latent Space Model With Position Clusters and Its Application in International Trade Network

The dynamic latent space model is widely used in analysing network data. It can provide useful visualization and interpretation of networks, as well as represent the inherent reciprocity and transitivity. In this paper, a dynamic latent space model with position clusters is proposed. The model extends the dynamic latent space model by incorporating latent position clustering and accounting for weighted networks. A fully Bayesian method with adaptive Markov chain Monte Carlo sampling is used to estimate the novel model. A purity‐based relabelling algorithm is proposed to resolve label switching. This model can be extended to analyse binary networks, count networks and sparse weighted networks. Finally, the model is used to analyse the product trade data of 54 countries from 2010 to 2019.