Latent Space Approaches to Social Network Analysis

Using the structure of social networks to map inter-agency relationships in public health services

“…Having interviewed our initial sample we modelled the partial network, using the concepts of latent position network models (Hoff, Raftery, &Handcock, 2002, Wasserman andFaust 1994) and latent position cluster models (Handcock, Raftery, & Tantrum, 2007). A model within two-dimensional latent social space was fitted to the initial interview data, which calculated the probability that individuals were connected.…”

Mixture of Experts Modelling with Social Science Applications

“…In particular, Hoff et al (2002) develop the idea of a latent social space and define the probability of a link between two actors as a function of their separation in the latent social space; this idea has been developed in various directions in Handcock et al (2007), Krivitsky and Handcock (2008) and Krivitsky et al (2009) in order to accommodate clusters (or communities) of highly connected actors in the network and other network effects. Airoldi et al (2008) develop an alternative latent variable model for social network data where a soft clustering of network actors is achieved; this has been further extended by Xing et al (2010) to model dynamic networks.…”

“…The latent position cluster model (LPCM) (Handcock et al 2007) develops the idea of the latent social space model (Hoff et al 2002) by extending the model to accommodate clusters of actors in the latent space. Under the latent position cluster model, the latent location of each actor is assumed to be drawn from a finite normal mixture model, each component of which represents a cluster of actors.…”

Adaptive Mixture Discriminant Analysis for Supervised Learning with Unobserved Classes

“…Among existing probabilistic social network models, we choose to consider the latent space model proposed by [23] and extended in [18] …”

“…, Z n have to be estimated for a fixed value of the latent space dimension p which can be chosen by crossvalidation or using a criterion such as BIC. Parameter estimation can be done by iterative likelihood maximization or MCMC techniques (see [23] for details). Recently, an extension of this model for the supervised classification problem has been proposed in [9].…”