-------------------------------------------------------------------------------
help for mmsrm
-------------------------------------------------------------------------------

Estimation of the parameters of a Multidimensional Marginally Sufficient Rasch 
> Model (MMSRM)

        mmsrm varlist [, partition(numlist) nodetails trace iterate(#) adapt
              method(mml/gee)]

        varlist is a list of two existing binary variables or more.

Description

    mmsrm estimates by marginal maximum likelihood (MML) or generalized
        estimating equations (GEE) the parameters of the Multidimensional
        Marginally Sufficient Rasch Model (MMSRM) defined by Hardouin and
        Mesbah (2004).  This model is an Item Response Model (IRM) with one
        or several latent traits.  This is a particular multidimensionnal
        extension of the Rasch model. In this model, the items are separated
        in Q groups and each group of items is linked to one and only one
        latent trait. Each group fits a Rasch model relatively to the
        corresponding latent trait, so the score computed in each group of
        item is a sufficient statistics of this latent trait (to a specific
        value of this score is associated only one value for the latent
        trait). The program allows computing the parameter of a MMSRM with
        less than 4 latent traits.  To improve the time of computing, the
        difficulty parameters are estimated in each unidimensional Rasch
        model and used as an offset variable to estimate the parameters of
        the distribution of the multidimensional latent trait. This model
        allows estimating the correlations between different latent traits
        measured by Rasch models.

Options

    partition allows defining the number of items relied to each latent
        trait. The sum of the numbers indicated in the numlist must be equal
        to the total number of items. The number of elements indicates the
        number of latent traits. The items are taken in the same order than
        this one of varlist.  By default, only one latent trait is assumed
        (Rasch model). The number of elements of partition must be inferior
        or equal to 3.

    method defines the estimation method between the marginal maximum
        likelihood (mml - by default) and the generalized estimating
        equations (gee).  You cannot estimate the parameters of a MMSRM with
        3 dimensions with the GEE method.

    nodetails deletes the details about the procedure.

    trace displays details about the estimation algorithm.

    iterate defines the maximal number of iterations of the estimation
        algorithm (30 by default).

    adapt allows using the adaptive quadrature if you use the MML method of
        estimation. This option is useless if you use GEE.

Example

                . mmsrm item1-item9 , part(4 5) trace method(gee)

                . mmsrm item1 item2 item3 item4 /*Rasch model*/

                . mmsrm c1-c9 , part(3 3 3) nodetails adapt iterate(10)

References

    Hardouin J.-B. and Mesbah M. Clustering binary variables in subscales
        using an extended Rasch model and Akaike Information Criterion.
        Communications in Statistics - Theory and Methods, 33(6), pp.
        1277-1294, (2004).

Author

    Jean-Benoit Hardouin, Regional Health Observatory (ORS) - 1, rue Porte
        Madeleine - BP 2439 - 45032 Orleans Cedex 1 - France.  You can
        contact the author at jean-benoit.hardouin@neuf.fr and visit the