ResearchTopics Publications In progress publications Communications
PackagesAnaQol Project PRO-online R Packages Online R-package
Life of the unitProjects Collaborations PhD thesis Traineeships Traineeships propositions Seminars
06 juin 2018
06 juin 2018
15 mai 2018
15 mai 2018
02 mai 2018
Updated20 juin 2018
The Stata module "Msp"
Msp allows realizing the Mokken Scale Procedure (MSP) defined by Hemker, Sijtsma et Molenaar (1995). This procedure selects sub-scales based on Loevinger H coefficients (1948). The results depend of a threshold "c" fixed by the user: the authors of the procedure recommand to use a "c" greater than 0.3. The kernel of the first sub-scale can be define. In the polytomous case, the Msp uses a Chi-deux test to test the significance of the coefficient, and not the Z test proposed by the authors of MSP.
Type "findit msp" or "ssc install msp" directly from your Stata browser.
Syntax (version 8.1)
msp varlist [, c(#) nodetails notest p(#) kernel(#) pairwise minvalues(#) nobon noadjust ]
This program requires an access to the following program(s):
- Displays, for each sub-scale, the Loevinger H index, and the Loevinger Hj indexes for each items.
- For polytomous items, msp uses a chi-deux test to test the nullity of the Loevinger H coefficients instead of the Z test as proposed by the authors of this procedure.
- c(#): Defines the value of the threshold c (0.3 by default).
- nodetails: Avoids to display the details of the algorithm
- notest: Avoids to test, for a new item, that Hjk>0 (The item with the greatest Hj is automatically selected)
- p(#): Defines the significance level of the tests
- kernel(#): Defines the kernel of the first sub-scale as the # first items of varlist
- pairwise: By default, all the individuals who have at least one missing values among all the items specified in varlist are deleted. This option allows computing the Loevinger H coefficients with all the possible individuals, for each pair of items.
- minvalues(#): Defines the minimum possible value for a Hjk coefficient of two items of a same scale (by default, fixed to 0).
- nobon: Avoids the Bonferroni corrections on the level of significance (the global level of significance is this one defined in the p option).
- noadjust: approximates the tests statistics like the MSP program (Molenaar et al. (2000)).
msp itemA1-itemA7, nobon
msp item*, c(.2) nodetails kernel(3)
msp itemA* itemB*,c(.65) pairwise notest