Files
Abstract
Recently, researchers and practitioners have shown increasing interest in the development and application of ideal-point Item Response Theory (IRT) models for constructing and scoring measures of constructs. This trend is largely due to theoretical and psychometric advantages presented by the ideal-point models over conventional dominance models for non-cognitive constructs. Applying ideal-point models requires researchers to first affirm the models’ dimensionality assumption aligned with data to be evaluated. It has been shown that violations of the dimensionality assumption will not only introduce serious biases in parameter estimation but may further result in misleading conclusions regarding relationships between constructs. However, due to the “extra-factor phenomenon” underlying the unfolding data structure, there exists a lack of techniques and consensus on ideal-point models’ dimensionality testing. The current study proposes a theory-based index to address this issue. With simulated data, the new index is compared with previously proposed numerical techniques and shows a number of advantages. As a result, the new approach is promising to help researchers to assess the dimensional structure of unfolding data without the need for datasets large enough to estimate the most common ideal point model, the generalized graded unfolding model (GGUM). Practices and implications for this new method will be discussed.