-------------------------------------------------------------------------------
help for msp
-------------------------------------------------------------------------------

Mokken Scale Procedure

        msp varlist [, nodetails kernel(#) c(#) notest p(#) pairwise
              minvalue(#) nobon noadjust]


        varlist is a list of two existing dichotomous or polytomous variables
              or more.

Description

    msp selects scales of dichotomous or polytomous items with the Mokken
        Scale Procedure (MSP) defined by Mokken (1971) and by Hemker and al.
        (1995).

Options

    nodetails does not display informations during the algorithm of
        selection.

    kernel defines a kernel for the first scale, defined as the first # items
        of varlist.

    c defines the level of the threshold c. The level of c is fixed, by
        default, at 0.3.

    notest selects the scales whithout tests among the Loevinger H
        coefficient (Hjk>0): the new item with the maximal value for Hj>c is
        always selected.

    p defines the significance level for the tests. The level of p is fixed,
        by default, at 0.05.

    pairwise does no more delete all the individuals who have one or more
        missing values among all the items of varlist but, only these ones
        who have, for each pair of items, one or more missing values to these
        two items.

    minvalue defines the minimum value of the coefficient Hjk for a pair of
        item composing a scale. This minimal value is fixed by default to 0.

    nobon does no more correct the significance level of the test by the
        Bonferroni correction.

    noadjust approximates the tests statistics like the MSP program (Molenaar
        et al. (2000)).

Remarks

    For detailed informations on the Mokken scale procedure, see Mokken
        (1971), Hemker and al. (1995) or the manual of the MSP software
        (2000).

    msp runs with dichotomous or polytomous items.

    msp necessitates the program loevh to compute.

Results

    r(dim) saves the number of built scales.

    r(selection) saves a vector which contains, for each item, the scale
        where it is selected (0 if the item is unselected).

    r(H#) saves the Loevinger's H coefficient of the #th scale.

    r(nbitems#) saves the number of items selected in the #th scale.

    r(scale#) saves the names of the items selected in the #th scale.

    r(scalenum#) saves the numbers of the items selected in the #th scale.

Example

        . msp itemA1-itemA7

        . msp itemA* itemB* , nodetails kernel(2)

        . msp item* , c(0.2) nob pairwise


References

    Hemker B. T., Sijtsma K. and Molenaar I. W., Selection of unidimensional
        scales from a multidimensional item bank in the polytomous Mokken IRT
        model, Applied Psychological Measurement, vol.19(4), 1995, pp.
        337-352.

    Mokken R. J., A theory and procedure of scale analysis. New-York/Berlin :
        De Guyter, 1971.

    Molenaar I. W., Sijtsma K. and Boer P. User's manual of MSP5 for Windows.
        Groningen (The Netherlands): iec ProGAMMA, 2000.

Also see

    help for loevh (if installed), help for mokken (if installed)

Author

    Jean-Benoit Hardouin, PhD, assistant professor
    EA 4275 "Team of Biostatistics, Clinical Research and Subjective Measures
        in Health Sciences"
    University of Nantes - Faculty of Pharmaceutical Sciences
    1, rue Gaston Veil - BP 53508
    44035 Nantes Cedex 1 - FRANCE
    Email:  jean-benoit.hardouin@univ-nantes.fr