Network Regression with Latent Communities

Extending Hoff's Additive and Multiplicative Effects Network Regression Model to allow for community dependent covariate coefficient estimation when communities are unknown.

Working on testing when censoring occurs in a network. If censoring has occurred, can we accurately predict who someone could possibly be friends with?

Latent communities often drive social connections. This makes learning these communities a challenging yet important problem to solve

Implementation and Highlights of this Work

• Developing MCMC with Metropolis Hastings steps to sample from this model

• Simultaneously learn community labels and community dependent coefficients

• Using data augmentation and post processing steps to make sampling feasible

• Real data applications