Network vector autoregression. (English) Zbl 1381.62256

In this manuscript, the authors introduce a network vector autoregressive (NAR) model to model some popular network structures including large-scale social networks. Two different types of stationarity are considered, when the network size is fixed and when the network size tends to infinity. For both cases the stationary conditions are obtained. The unknown parameters of the NAR model are estimated by the ordinary least squares method and the asymptotical normality of the obtained estimators is proved. Some potential applications to existing structures are discussed. At the end of the paper, some numerical simulations are provided to check the finite sample performance of the proposed methodology.


62M10 Time series, auto-correlation, regression, etc. in statistics (GARCH)
62J05 Linear regression; mixed models
Full Text: DOI Euclid