|
Psychometric Research & Development Lab |
|
To contact us: School of Psychology J.S. Coon 137 Georgia Institute of Technology 654 Cherry Street Atlanta, GA 30332 Phone: 404-385-2985 Fax: 404-894-8905 Email: robertslab@hotmail.com |
|
J.S. Roberts & V.M. Thompson This project implements and compares a marginal Bayesian estimation method for GGUM parameters to a fully Bayesian (Markov Chain Monte Carlo) estimation method. We hypothesize that both methods will reduce data demands and improve estimation accuracy, whereas the marginal approach should be much faster computationally. |
|
H. Shim, V.M. Thompson, J.D. Gordon, Z. Morford, S. Bhanjee, C. Bowden, A. Monteiro, M. Dolphyn, A. Skeete, Z. Blanton, & J.S. Roberts In this project, item-level data are gathered using a questionnaire designed to measure attitude toward abortion. Respondents are repeatedly measured across an approximate 3-week span. Responses will be used to test several new techniques including alternative characteristic curve linking methods and the generalized graded unfolding model for repeated measures. |
|
J.S. Roberts & H. Shin The GGUM2004 software package is a user-friendly, Windows-based system to estimate parameters of the Generalized Graded Unfolding Model (GGUM) and diagnose model performance. It is continuously maintained by the program authors and periodically revised to incorporate new features. The software is freely distributed and can be downloaded from www.psychology.gatech.edu/unfolding. |
|
W. Cui & J.S. Roberts This project develops and implements a version of the GGUM for repeated measures situations in which either the same test or alternate forms of a test (with embeded anchor items) are given repeatedly to the same cohort of individuals. The parameters are estimated with a fully Bayesian (MCMC) approach. The data demands required for accurate parameter estimation are studied and examples of model applications are explored. |
|
A Comparison of Marginal and Fully Bayesian Estimation of Generalized Graded Unfolding (GGUM) Parameters |
|
GGUM2004 Program Development and Maintenance |
|
Attitude Data Collection Project |
|
The Generalized Graded Unfolding Model for Repeated Measures (GGUM-RM) |
Current Projects |
|
H. Shim & J.S. Roberts This project will explore item-level responses to repeated test administrations using a relatively new family of monotone item response theory models that can be used when respondents are measured over time with alternate test forms containing embedded common items. The model will account for dependencies that arise due to repeated examination of the same individuals. |
|
Exploring Item-Level Responses in the Early Childhood Longitudinal Study |
|
J.S. Roberts & X. Huang This project looks at mixtures of the Generalized Graded Unfolding Model (GGUM) along with mixtures of a multinomial model of random responses to explain variability in responses to attitude questionnaires. The resulting mixture model will assess the probability that a given subject is responding randomly to questionnaire items as opposed to responding in a systematic manner as defined by the GGUM. The probability of responding randomly can be used as an index of person-level fit to the model. |
|
Mixtures of the Generalized Graded Unfolding Model (GGUM) that Include Random Responders |