The FMM Procedure

Basic Features

The FMM procedure estimates the parameters in univariate finite mixture models and produces various statistics to evaluate parameters and model fit. The following list summarizes some basic features of the FMM procedure:

  • maximum likelihood estimation for all models

  • Markov chain Monte Carlo estimation for many models, including zero-inflated Poisson models

  • many built-in link and distribution functions for modeling, including the beta, shifted t, Weibull, beta-binomial, and generalized Poisson distributions, in addition to many standard members of the exponential family of distributions

  • specialized built-in mixture models such as the binomial cluster model (Morel and Nagaraj, 1993; Morel and Neerchal, 1997; Neerchal and Morel, 1998)

  • acceptance of multiple MODEL statements to build mixture models in which the model effects, distributions, or link functions vary across mixture components

  • model-building syntax using CLASS and effect-based MODEL statements familiar from many other SAS/STAT procedures (for example, the GLM, GLIMMIX, and MIXED procedures)

  • evaluation of sequences of mixture models when you specify ranges for the number of components

  • simple syntax to impose linear equality and inequality constraints among parameters

  • ability to model regression and classification effects in the mixing probabilities through the PROBMODEL statement

  • ability to incorporate full or partially known component membership into the analysis through the PARTIAL= option in the PROC FMM statement

  • OUTPUT statement that produces a SAS data set with important statistics for interpreting mixture models, such as component log likelihoods and prior and posterior probabilities

  • ability to add zero-inflation to any model

  • output data set with posterior parameter values for the Markov chain

  • high degree of multithreading for high-performance optimization and Monte Carlo sampling

The FMM procedure uses ODS Graphics to create graphs as part of its output. For general information about ODS Graphics, see Chapter 21: Statistical Graphics Using ODS. For specific information about the statistical graphics available with the FMM procedure, see the PLOTS options in the PROC FMM statement.