网络因子模型建模方法的回顾、评述与展望OACHSSCD

In recent years,research on network dimension reduction techniques based on network factor models and related applications have been emerging continuously.Based on a brief review of the node factor model,this paper focuses on systemati-cally sorting out the modeling methods of the network factor model considering edges.According to the static and dynamic charac-teristics of the common factors,the matrix value and vector value structure of the common factors,as well as the exogeneity and en-dogeneity of the network structure,the network factor models considering edges are classified into four categories.From the per-spectives of econometric analysis such as model setting and model estimation,the related studies are classified,summarized and reviewed.It can not only achieve dimensionality reduction of node variables but also of network connections.Strengthening the identifiable analysis of models,paying attention to the robustness of model estimates and the"curse of dimensionality",remains a key focus for future research.