Multivariate Normal Mixture Models and Mixtures of Generalized Linear Mixed Models Including Model Based Clustering

Contains a mixture of statistical methods including the MCMC methods to analyze normal mixtures. Additionally, model based clustering methods are implemented to perform classification based on (multivariate) longitudinal (or otherwise correlated) data. The basis for such clustering is a mixture of multivariate generalized linear mixed models.



26/11/2010 TO DO: fitted.GLMM_MCMC not yet implemented for situations when there are only fixed effects in the model

14/06/2010 TO DO: In GLMM_Deviance.cpp Commands: *marg_ll_iP += *w_k * AK_Basic::exp0_AK(loglik_k); (around line 239) *marg_ll_iP = AK_Basic::log0_AK(*marg_ll_iP); (around line 252) In the case of the poisson response, it would be desirable to add log(y!) to loglik_k before exp-it
and subtract it again after log the mixture sum. With higher counts, loglik_k is quite considerably negative which leads to exp(-Inf) = 0 and then upon log again, we get -Inf log-likelihood where -Inf is artificial due to imprecise calculation with numbers close to zero.

--> more or less solved on 03/12/2010

====================================================================================================================== 17/05/2018 (version 5.1): Minor changes in plotProfiles.R and in NMix_RJMCMCcombine.cpp

15/03/2017 (version 5.0): Registration of C routines used.

11/10/2016 (version 4.4): Formal correction of some header files (e-mail B. Ripley from 18/09/2016).

02/06/2016 (version 4.3): Argument type added to cbplot function.

01/08/2015 (version 4.2): Bug in getProfiles corrected that was incidentally created in version 4.1.

28/07/2015 (version 4.1): Options to parallel computation added (on 29/06/2015).

26/03 - 02/04/2015 (version 4.0): Possibility of having weights dependent on a categorical covariate implemented in NMixMCMC and related functions.

06/02/2015 (version 3.10): NAMESPACE updated by explicit export of some S3 methods.

05/02/2015 (version 3.9): CITATION file updated (desc replaced by meta).

01/09/2014 (version 3.7): Bug in GLMM_MCMCdata corrected (it appeared when two or more discrete responses of the same type, e.g., binomial, were included in the model).

CITATION file updated (methodological papers removed from there).

21/08/2014 (version 3.7): NMixCluster function added (currently only method for class GLMM_MCMC).

06/08/2014 (version 3.6-1): Update of index.html in /inst/doc directory, validation using

14/05/2014 (version 3.6): Minor changes related to the final JSS submission.

22/09/2013 (version 3.5): Gauss-Hermite quadrature inside the fitted.GLMM_MCMC function implemented using the functions from the 'fastGHQuad' package. Arguments x2, z2 are no more needed.

16/08/2013 (version 3.4-2): Function NMixRelabel.GLMM_MCMClist allows for a joint processing of two chains (jointly = TRUE).

06/08/2013 (version 3.4-1): Functions rWishart and dWishart renamed to rWISHART and dWISHART to avoid conflicts with rWishart in package stats.

05/08/2013 (version 3.4): Most dependencies moved from "Depends" to "Imports" in the DESCRIPTION file and solved via importFrom in the NAMESPACE.

18/07/2013 (version 3.2-4): Changes of names of some components inside the objects of classes NMixMCMC and GLMM_MCMC: comp.prob2 --> comp.prob_b comp.prob3 --> comp.prob

poster.comp.prob1 --> poster.comp.prob_u poster.comp.prob2 --> poster.comp.prob_b poster.comp.prob3 --> poster.comp.prob

quant.comp.prob1 --> quant.comp.prob_u quant.comp.prob2 --> quant.comp.prob_b quant.comp.prob3 --> quant.comp.prob

08/02/2013: Support for parallel computation originally provided by snowfall/snow packages changed to parallel package.

01/05/2012: All pdf vignettes moved to my personal webpage (to decrease the size of the package which must be stored on CRAN).

18/12/2011: GLMM_MCMC creates by default two parallel chains and calculates also PED, support for parallel computation on multicore CPUs using snowfall/snow packages added.

20/01/2011: beta in GLMM_MCMC function (and all related) was renamed to alpha to be in agreement with papers.

29/11/2010: Calculation of posterior quantiles for estimated component probabilities added to NMixRelabel functions (both NMixMCMC and GLMM_MCMC version).

08/12/2009 Bug in computation of weights for the EB estimate of random effects in GLMM_longitPred_nmix_gauss.cpp fixed.

22/11/2009 Function NMixClust renamed to NMixPlugDA.

06/11/2009 Bugs in GLMM_updateRanEf.cpp found and fixed:

  • eta_random was computed incorrectly as Z %% bscaled and not as Z %% b, where b = shift + diag(scale) %*% b
  • pointers in the final part of the code did not shift correctly

Reference manual

It appears you don't have a PDF plugin for this browser. You can click here to download the reference manual.


5.3 by Arnošt Komárek, a year ago

Browse source code at

Authors: Arnošt Komárek <[email protected]>

Documentation:   PDF Manual  

Task views: Cluster Analysis & Finite Mixture Models, Survival Analysis

GPL (>= 3) license

Imports graphics, stats, splines, fastGHQuad, mnormt, parallel, coda

Depends on colorspace, lme4

Suggested by icensBKL, latrend, miscF.

See at CRAN