Current Projects

Attitude Data Collection Project

E. Williams, J. Sparks,  & J. S. Roberts
In this project, item-level data are gathered using questionnaires designed to measure attitudes toward abortion, gun control and politics. Respondents are repeatedly measured across an approximate 3-week span. Responses will be used to test several new psychometric techniques in the domain of unfolding item response theory.

Applications of the Multidimensional Graded Unfolding Model (MGGUM): The Emotion Faces Study

J. Sparks, M. Dolphyn & J. Roberts
This study examines judgments about the similarity of facial emotions in photographs and creates a multidimensional space of that contains those emotions. Additionally, the project also examines judgments of similarity to identify which emotions a respondent has recently experienced. Respondents are located in the multidimensional space near those emotions that they have recently felt. The construction of the multidimensional space will be performed using both classical multidimensional scaling techniques along with new multidimensional generalized graded unfolding model (MGGUM).

Applications of the Multidimensional Graded Unfolding Model (MGGUM): The Physical Attraction Study

M. Barrett, J. Roberts & X. Xiong
This project examines ratings of physical attractiveness of computer generated female models. The ratings are used to locate the models in a multidimensional space. Respondents are subsequently located in the same space by fitting different forms of regression models dictating different types of preferences. Both classical multidimensional scaling methods and the new multidimensional generalized graded unfolding model will be used to develop the physical attraction space, and these spaces will be compared.

Alternative Estimation in the Multidimensional Generalized Graded Unfolding Model (MGGUM)

D. King & J. Roberts
This project implements and investigates alternative methods of estimating parameters in the multidimensional generalized graded unfolding model. Methods include marginal maximum likelihood (MML) and the Metropolis-Hastings Robbins-Monro (MH-RM) algorithms. The accuracy of these methods will be contrasted and a general program will ultimately be developed for public use.

Constrained Estimation in the Multidimensional Generalized Graded Unfolding Model (MGGUM)

E. Williams & J. Roberts
This study constrains item location parameters in the multidimensional generalized graded unfolding model to values derived from alternative methods in an effort to minimize data demands. Location estimates derived from multidimensional scaling of pairwise similarity judgments are contrasted with those developed from detrended correspondence analysis of single stimulus judgments, and each type are included into the MGGUM estimation algorithm.

Dimensionality Assessment in the Multidimensional Generalized Graded Unfolding Model (MGGUM)

E. Williams & J. Roberts
This project will develop methods to assist practitioners with dimensionality assessment in the multidimensional generalized graded unfolding model. Unfolding models are unique in that they incorporate nonmonotonic, single peaked item response functions. In contrast, typical dimensionality assessment techniques assume monotonic item response functions. Therefore, such techniques are not applicable to the unfolding realm. The goal of this study is to develop tools to estimate the dimensionality of unfolding item responses that are both practical and accurate.

Dimensionality Assessment in the Multidimensional Generalized Graded Unfolding Model (MGGUM)

R. Meyer & J. Roberts
This study examines the effectiveness of MGGUM comparisons to determine the appropriate number of dimensions reflected in unfolding item responses. Both statistical tests and model fit criteria are used to determine the optimal model when MGGUMs with varying dimensionality are fit to simulated data with known dimensionality. The results of this study will determine if dimensionality assessments via model fit will be useful to measurement practitioners.