A note on estimating network dependence in a discrete choice model (Q1782133)

This paper studies estimating of network dependence in a discrete choice model in the context of a latent spatial auto-regressive model. Network dependence is considered under a principled binary regression method. A pseudo maximum likelihood method is adopted to estimate the spatial auto-correlation of the network. \(n\) nodes are structured into different pairs and each pair contains two nodes \(i\) and \(j\), leading to \(n(n-1)/2\) pairs. Based upon Taylor's expansion, a log likelihood function is approximated for each pair of nodes. Selected pairs for objective functions have been summed assuming certain independence. The consistency of approximated paired maximum likelihood estimator is studied and a numerical simulation for Sina Weibo is presented.