Package: EMfit 0.2.1

EMfit: An Infrastructure for Latent Variable Model Fitting using EM Algorithms

The package contains the boilerplate code for estimating latent variable models by maximum marginal likelihood using EM algorithms. The code is generic and thus not optimised for particular models. The package supports fitting latent class or finite mixture models using simple EM algorithms and (to some degree) latent trait or continous mixture models using Monte Carlo EM algorithms. Only importance sampling is supported for MCEM. However, it implements automatic sample size adjustment using the approach of Caffo Jank and Jones.

Authors:Martin Elff