Presentation

HETMIXLIN is a program for estimation of linear mixed models with a mixture of distribution for the random effects. This kind of models is an extension of linear mixed models for heterogeneous data stemmed from various latent classes of evolution. For a given number of latent classes G, HETMIXLIN allows to estimate the parameters of the G profiles of evolution accounting for time-dependent covariates and to classify subjects according to the profile for which they have the highest posterior probability to belong. Parameter estimation is performed by a Maximum Likelihood method : the observed log-likelihood is maximized directly using a modified Marquardt algorithm.

In the parameter file, the user specifies the number of components of mixture and can specify for each covariate if the parameter is random and/or the parameter is specific to each component of mixture. 

HETMIXLIN.pdf	user's guide
HETMIXLIN.f	source file in Fortran90
HETMIXLIN.inf	parameter file described below
schoolgirls.txt	example of a data file
HETMIXLIN	executable version of HETMIXLIN.f for Unix

Results

The output file for the estimations contains the main specifications of the model, the convergence status, the final log-likelihood, the Akaike criterion, the Bayesian Information Criterion, parameter estimates with their standard errors, the Wald test statistic and the 95% Confidence Interval of the estimate

The output file for the posterior probabilities contains a line per subject with the identification number, the posterior probability to belong to each of the G components and the final classification variable.

An example of each output file is described in the user's guide.

Data file

This ASCII file contains the data set and must be given in the following format:

The data file is constitued in lines. It contains for each subject:

• identification number
• number of measures n_i
• n_i vector of response
• n_i vector of measures for the covariate 1
• n_i vector of measures for the covariate 2
• ...
• n_i vector of measures for the covariate K

Remarks:

• the intercept must be contained in the covariates
• even if the value does not change, the covariate must be specified in a n_i vector
• the name of the data file is given in the parameter file

Parameter file:

This file contains all the information needed for a complete specification of the model :

• name of the data file
• name of the output file for the estimations
• procedure title
• number of subjects
• number of components in the mixture (with below if needed initial values of the probabilities and name of the posterior probability file)
• number K of covariates in the data file
• indicator of their presence in the model
• indicator of a random parameter for each covariate included in the model
• indicator of a mixture parameter for each covariate included in the model
• initial values for fixed effects
• indicator of the random-effect covariance matrix structure
• initial values for the variance-covariance parameters of the random effects
• initial value for the variance of the independant gaussian errors

Each line of information to specify is preceded by a line summing up what is asked.

References

Proust Cécile and Jacqmin-Gadda Hélène
Estimation of linear mixed models with a mixture of distribution for the random effects.
Comput Methods Programs Biomed. 2005 May;78(2):165-73.

Authors

Cécile Proust
Hélène Jacqmin-Gadda

Inserm U897
146 rue Léo Saignat
33076 Bordeaux Cedex
France

Contact

E-mail: Daniel.Commenges@isped.u-bordeaux2.fr.
We are interested in feed-back but can not guarantee support.

Licence

The authors are not in business of writing software packages and can not garantee that the program is error free. The main purpose of this program is for our own use in analyzing data. It is distributed free of charge for use by non-programmers and for programmers who would like to extract parts and possibly modify the program.

• HETMIXLIN (Unix version)
• HETMIXLIN (MS Windows version)