On 31.08.2017 16:35, Valery Topinsky wrote:
Hello.
After reading your "Nonparametric weighted stochastic block models" paper and trying weighted NSBM model. I'm trying to find inferred distribution (mixed) for weights (edge-covariates). How can I get them from the BlockState object? I need it to repeat your graphical analysis on different weight transforms from the paper. I guess somehow I can restore the "the marginal covariate distribution between groups r and s" from resulted object.
You need to compute the marginal distribution yourself, using Eq. 15 together with the appropriate marginal covariate distribution, e.g. Eq. 51 for the signed weights. This is not yet implemented automatically in BlockState. Best, Tiago -- Tiago de Paula Peixoto <tiago@skewed.de>