Calibrating the Test of Relational Reasoning: New Information from Oblique Bifactor Models
Relational reasoning is believed to be an essential construct for studying higher education learning. Relational reasoning is defined as an ability to discern meaningful patterns within any stream of information. Nonetheless, studies of relational reasoning are limited by the psychometric structure of the construct. For many instances, the composite nature of relational reasoning has been described as a bifactor structure. Bifactor models limit possibilities for studying the inner structure of composite constructs by demanding orthogonality of latent dimensions. Such assumption severely limits the interpretation of the results when it is applied to psychological constructs. However, over the last ten years, advances in the fields of Rasch measurement led to the development of the oblique bifactor models, which relax the constraints of the orthogonal bifactor models. We show that the oblique bifactor models exhibit model fit, which is superior compared to the orthogonal bifactor model. Then, we discuss their interpretation and demonstrate the advantages of these models for investigating the inner structure of the test of relational reasoning. The data is a nationally representative sample of Russian engineering students (N = 2036).