| MSP - Nonparametric IRT Scaling |
W. Molenaar, University of Groningen, Netherlands
P. Debets, University of Amsterdam, Netherlands
K. Sijtsma, Utrecht University, Netherlands
B.T. Hemker, CITO Arnhem, Netherlands
MSP, Mokken scale analysis for polytomous items, scales test
item responses using nonparametric cumulative item response
theory. It can test scalability of a given scale or construct one
or more unidimensional scales from an item pool. Persons are
ordered by scale score and measurement quality is assessed.
MSP can analyze either dichotomous or polytomous data, and can
be used for:
- Stepwise construction of one or more unidimensional
scales from a pool of items;
- Evaluation of model fit of a given scale, including the
assessment of its reliability and suggestions for removal
of misfitting items; and
- Detection of the extent to which the answer patterns
follow the cumulative Guttman scalogram pattern, using
Loevinger's H-values and associated significance tests.
The latest version incorporates many new
features and improvements:
- An improved definition of the scalability coefficients;
- Computation of the frequency distribution and
corresponding statistics of the scale score, and the
number of Guttman errors per person;
- An option to analyze and compare several subgroups in one
run;
- Checks the assumptions underlying the models for
monotone homogeneity and double monotonicity; and
- Estimates the reliability of a
test score.
The scaling procedures of the program can handle up to 100
polytomous items with a maximum of 10 ordered categories within a
given range (900 item steps). Persons with values outside this
range are deleted. Data for up to 32,000 persons can be analyzed. No missing data
is allowed.
Related Books
System Requirements
Windows 95, 98, NT, ME, 2000, XP
Demonstration Version
The demo version of MSP will run only the sample files that come with it. In all other
respects it is identical to the regular version.
