Tools for Discrete Multivariate Mixed Membership Models
Description
Fits mixed membership models with discrete multivariate data (with
or without repeated measures) following the general framework of Erosheva et al
(2004). This package uses a Variational EM approach by approximating the
posterior distribution of latent memberships and selecting hyperparameters
through a pseudo-MLE procedure. Currently supported data types are
Bernoulli, multinomial and rank (Plackett-Luce). The extended GoM model with fixed stayers from Erosheva et al (2007) is now also supported. See Airoldi et al (2014) for other examples of mixed membership models.