bmggum: An R Package for Bayesian Estimation of the Multidimensional Generalized Graded Unfolding Model With...

Area of Science:

• Psychometrics
• Statistical Modeling
• Behavioral Science

Background:

• Ideal point models are increasingly used for noncognitive constructs, outperforming traditional dominance models.
• The generalized graded unfolding model (GGUM) is a popular choice, but existing software is limited to unidimensional applications.
• Current GGUM implementations can produce unreliable item parameter and standard error estimates.

Purpose of the Study:

• To introduce the open-source bmggum R package for estimating unidimensional and multidimensional GGUM.
• To address limitations in existing GGUM software regarding dimensionality and parameter estimation accuracy.
• To provide a robust Bayesian framework for analyzing noncognitive constructs.

Main Methods:

• Development of the bmggum R package utilizing a fully Bayesian approach.
• Implementation of capabilities for parameter stabilization, person covariate incorporation, and constrained model estimation.
• Inclusion of fit diagnostics, convergence metrics, and missing data handling.

Main Results:

• The bmggum package successfully estimates both unidimensional and multidimensional GGUM.
• The Bayesian approach provides stable parameter estimates and reliable standard errors.
• The package effectively handles missing data and offers comprehensive diagnostic tools.

Conclusions:

• The bmggum R package offers a significant advancement for researchers analyzing noncognitive constructs.
• It overcomes the dimensionality limitations of previous GGUM implementations.
• This tool facilitates more accurate and flexible modeling of complex behaviors.