| DIFPACK - Dimensionality-Based DIF Analysis Package |
Includes SIBTEST, POLY-SIBTEST, Crossing SIBTEST, DIFCOMP, and DIFSIM
The William Stout Institute for Measurement
These products were originally announced as separate DOS programs. Due to popular demand they are now only available in a package with A 32-bit Windows interface. Printed documentation that is shipped with the CD includes instructions, a brief description of each of the programs, and copies of nine relevant journal articles to provide a technical background on the topic.
Differential Item Functioning
DIFCOMP computes the theoretical amount of DIF/DBF for an item
or item bundle for a given two-dimensional IRT model known to
produce DIF. It computes the theoretical b index, which
measures the amount of DIF/DBF present, for suspect
multidimensional items or item bundle given a particular pair of
multidimensional ability distributions for the focal and
reference groups and the two-dimensional item response funtions
of the studied item(s). The user specifies parameters of the
bivariate normal distribution of the target (intended) ability
and the secondary ability for two populations of examinees,
multidimensional logistic parameters for the DIF-displaying
item(s) with respect to each ability, whether the amount of
unidirectional or crossing DIF/DBF is to be computed, and the
desired precision of the integration. Differences in expected
item or item bundle scores are computed at multiple locations on
the target ability continuum and averaged over ability to compute
an approximate theoretical b index. DIFCOMP also permits
assessment of DIF amplification (i.e., amount of DBF) by allowing
the user to enter parameters for multiple items and compute the
combined theoretical b index for the resulting bundle of
items.
DIFSIM - Multidimensional Differential Item Functioning
DIFSIM generates item responses for focal and reference
populations of examinees based on Shealy-Stout’s
multidimensional model for DIF. The user specifies the marginal
bivariate distributions of the target ability and each of up to
10 secondary abilities, logistic parameters for all items with
respect to each dimension with item parameters supplied by the
user or randomly generated, number of examinees in each
population, and number of DIF and "valid" items
(assumed strictly unidimensional). An output file for each
population containing the examinees’ item responses is made.
The generated data can then be used, e.g., to estimate the
SIBTEST theoretical b index of DIF/DBF magnitude for an
item or items. DIFCOMP is useful with DIFSIM: it computes the
theoretical value of b for a given multidimensional DIF
model like that used in DIFCOMP to simulate DIF/DBF data.
SIBTEST - Differential Item/Bundle Functioning
SIBTEST implements a nonparametric
statistical method of assessing DIF/DBF (differential item
functioning/differential bundle functioning) in an item or bundle
of items based on Shealy-Stout’s (1993) multidimensional
model for DIF. Matching on target ability is done by matching on
total test score or other score believed to validly measure the
target ability (i.e., the dimension best measured by reported
test score), such as total score with certain items removed that
are believed to be DIF-producing. With a flexible front-end
program, the user specifies the particular DIF/DBF hypothesis
tests to be performed—including which item or items will be
tested, which population (reference, focal, or either) DIF/DBF
will be tested to occur against, and which items will be used to
construct the matching score. SIBTEST uses a recently developed
nonlinear regression correction procedure to match examinees,
which has demonstrated particularly good effectiveness in
controlling the false flagging of non-DIF items.
Crossing SIBTEST
Crossing SIBTEST tests for crossing DIF, the
condition in which reference and focal group differences in
expected score at each fixed target ability value on an item or
item bundle between two populations change sign at some point on
the target ability scale. Analogous to SIBTEST, this program uses
a regression correction technique to appropriately match
examinees from the two populations, and computes the expected
score differences for a suspect item or item bundle at varying
ability levels. Fundamental to the procedure is its capacity to
identify the location of the crossing point (e.g. the target
ability level at which the expected score difference changes
sign). Output includes an estimate of the theoretical b
crossing-DIF index, the point at which crossing occurs, the
relative amount of DIF/DBF occurring against each population
(necessary because crossing DIF results in DIF against both
populations), and the results of the DIF/DBF significance test.
The procedure is also sensitive to noncrossing DIF/DBF, although
it differs some in algorithmic detail from regular SIBTEST.
Polytomous SIBTEST
Polytomous SIBTEST handles polytomously
scored item responses. This extension of SIBTEST is
straightforward, except for some changes to account for the
additional score categories in the matching criterion. It uses
revised reliability estimates in performing the regression
correction, with the regression correction being a generalization
of the original linear regression correction.
System Requirements
Windows 95, 98, NT, ME, 2000, XP
DIFPACK Manual
The manual for DIFPACK includes instructions for installing the program, technical descriptions of each component program, a step-by-step guide to using the program, and copies of references to provide a theoretical background.
View technical descriptions and reference list here.
|
Order Options and Details
|
 |
| |
Special Introductory Bundle
Purchase DIFPACK and DIMPACK at the same time, and receive a 25% discount:
Products
Your Account
Your cart
