The package LMest is a framework for specifying and fitting Latent (or Hidden) Markov (LM) models for the analysis of longitudinal continuous and categorical data. Covariates are also included in the model specification through suitable parameterizations.
Francesco Bartolucci [aut,cre], Silvia Pandolfi [aut], Fulvia Pennoni [aut], Alessio Farcomeni [ctb], and Alessio Serafini [ctb]
Maintainer: Francesco Bartolucci <francesco.bartolucci@unipg.it>
Different LM models are estimated through specific functions requiring a data frame in long format. Responses are mainly categorical, the functions referred to continous responses are specified with Cont.
The functions are the following:
lmestFunction to estimate LM models for categorical responses generating the following classes:
LMbasic-class for the basic LM model without covariates.
LMmanifest-class for the LM model with covariates in the measurement submodel.
LMlatent-class for the LM model with covariates in the latent model.
lmestContFunction to estimate LM models for continuous outcomes generating the following classes:
LMbasiccont-class for the basic LM model for continuous responses without covariates.
LMlatentcont-class for the LM model for continuous responses with covariates in the latent model.
lmestMixedFunction to estimate Mixed LM models for categorical responses with discrete random effects in the latent model generating the following class:
LMmixed-class for the mixed LM model.
lmestMcFunction to estimate Markov Chain models for categorical responses generating the following classes:
MCbasic-class for the Markov Chain (MC) model without covariates.
MCcov-class for the MC model with covariates.
Maximum likelihood estimation of model parameters is performed through the Expectation-Maximization algorithm, which is implemented by relying on Fortran routines.
Model selection is provided by lmest and lmestCont functions. In addition, function lmestSearch allows us to deal with both model selection and multimodality of the likelihood function. Two main criteria are provided to select the number of latent states: the Akaike Information Criterion and the Bayesian Information Criterion.
Prediction of the latent states is performed by the function lmestDecoding: for local and global decoding (Viterbi algorithm) from the output of functions lmest, lmestCont and lmestMixed.
The package allows us to deal with missing responses, including drop-out and non-monotonic missingness, under the missing-at-random assumption.
Standard errors for the parameter estimates are obtained by the function se through exact computation of the information matrix or by reliable numerical approximations of this matrix.
The print method shows some convergence information, and the summary method shows the estimation results.
The package also provides some real and simulated data sets that are listed using the function data(package = "LMest").
Bartolucci, F., Pandolfi, S. and Pennoni, F. (2017). LMest: An R Package for Latent Markov Models for Longitudinal Categorical Data, Journal of Statistical Software, 81, 1-38, doi:10.18637/jss.v081.i04.
Bartolucci, F., Farcomeni, A. and Pennoni, F. (2013). Latent Markov Models for Longitudinal Data, Chapman and Hall/CRC press.
Bartolucci, F., Farcomeni, A., and Pennoni, F. (2014). Latent Markov models: A review of a general framework for the analysis of longitudinal data with covariates (with discussion). TEST, 23, 433-465.
lmest, lmestCont, lmestMc, lmestMixed, LMmixed-class, LMbasic-class,LMbasiccont-class, LMlatent-class,LMlatentcont-class, LMmanifest-class