SHORT COURSE FOR MEASUREMENT AND RESEARCH PROFESSIONALS
The Department of Measurement, Statistics & Evaluation
at the University of Maryland, College Park will offer a
short course on "Analyzing Unfolding Item Response Theory
Models with GGUM2004 Software" on August 12-13. Responses
to many Likert-style or Thurstone-style attitude
questionnaires can typically be modeled with item response
theory (IRT). Over the last decade, several researchers
have argued that these data are often more consistent with
unfolding models rather than traditional cumulative models.
Cumulative IRT models suggest a "more is better"
orientation that works well for constructs like
proficiency, where we expect more proficient respondents to
get higher item scores on an achievement test relative to
less proficient individuals. In contrast, unfolding models
suggest that respondents will endorse attitude
questionnaire items to the extent that their opinions are
conceptually close to the content of those items. This
suggests an "ideal point" orientation. Roberts, Donoghue
& Laughlin (2000; Applied Psychological Measurement) have
built this notion into a family of IRT models. The most
general member of the model family is called the
Generalized Graded Unfolding Model (GGUM). The GGUM is
appropriate for analyzing unidimensional item responses
from typical attitude, satisfaction or preference rating
scales that are built in either the Thurstone of Likert
tradition.
The short course will consist of a didactic component
along with a computer-based, "hands-on" component. The
didactic component will assume minimal knowledge of IRT
models. It will review necessary concepts from traditional
IRT models for cumulative responses and demonstrate how the
GGUM family of unfolding models can be built from these
traditional models. Participants will learn the
characteristics of alternative models in the GGUM family,
and more importantly, they will learn to apply them.
Course topics will include:
- Writing questionnaire items in the Thurstonian tradition
- Selecting unidimensional items from an existing data set
- Estimating GGUM parameters
- Item and model fit assessment
- Linking parameters from alternative GGUM analyses
- Examining differential item functioning in the context of
the GGUM
The hands-on component will introduce participants to
the new GGUM2004 computer program. This is a Windows-based
program for estimating parameters in the GGUM, assessing
item and model fit, and graphically diagnosing model
performance. Participants will analyze both simulated and
real data sets that demonstrate particular features
addressed in the didactic component. Desktop computers and
appropriate software will be provided. Each participant
will receive a prerelease (beta) version of the GGUM2004
software program in CD format.
The short course will be conducted by Dr. James
Roberts on August 12-13 at the University of Maryland,
College Park. The registration fee is $275 and includes
all printed and electronic training materials, computer
access, parking and continental breakfasts. More
information about the short course and registration
procedure is available at:
http://www.education.umd.edu/EDMS/tutorials/scourse/unfold.pdf
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
. http://jse.stat.ncsu.edu/ .
=================================================================
