SMABS 2004 Jena University
SMABS 2004 Home Organization About Jena Sponsors Links Imprint / Contact SMABS Home

European Association of Methodology

Department of methodology and evaluation research

Jena University

Contributions: Abstract

Introducing GGUM2004: A Windows-based program to implement item response models for unfolding

James Roberts
University of Maryland

GGUM2004 is a freeware program that estimates parameters and provides pertinent diagnostics for the generalized graded unfolding model (GGUM). The GGUM is a unidimensional IRT model that unfolds polytomous responses to standard rating scales typically used in attitude and personality measurement. The most general form of the model can be constrained to provide up to seven alternative models. These include models in which item discrimination parameters and/or the subjective response category thresholds are constant across items.

The GGUM2004 program implements a marginal maximum likelihood (MML) solution for item parameters along with an expected a priori (EAP) solution for person parameters. Simulation studies have shown that parameter recovery with this approach is very good for the most general model when there are approximately 750-1000 respondents and 15-20 polytomous items. Sensitivity to the standard normal prior distribution is generally negligible except when estimating person parameters for the most extreme respondents.

GGUM2004 supercedes its predecessor, GGUM2000, and offers several new advantages. First, the GGUM2004 program implements a graphical user interface which enables the user to build a command file, run the computational component of the program, and examine results. Second, the new program will allow for randomly missing responses and responses to items with variable numbers of categories. Third, the program implements new item fit statistics that perform relatively well as compared to traditional chi-squared approaches to item fit. Finally, the GGUM2004 offers a large variety of real-time graphics that help users interpret model results and diagnose item and model fit.

In this software presentation, the GGUM2004 program will be demonstrated using data from both attitude and personality domains.