Full Program »
On the Roles of Conceptual Definitions and Construct Theories in Humanistic MeasurementProponents of Rasch modeling value its commitment to meet the mathematical criteria of fundamental measurement – i.e., measurement that is conjointly additive and generally objective. The properties of fundamental measurement are also characteristic of many archetypal measures in physics, and it may be argued that proponents of Rasch modeling in the social sciences wish to emulate the mathematical and structural properties of physical measures.
However, I contend that the properties and affordances associated with fundamental measurement in the physical sciences are adhered to not because of scientists’ commitments to fundamental measurement as a measurement model, but instead because of a commitment to certain physical theories. I.e., it is the physical theory that determines the structure of the measurement model (for instance, fundamental measurement), and not merely a commitment to a certain measurement model per se.
I consider as a case study the attempt certain Rasch proponents make to adhere to the mathematical requirements of fundamental measurement by measuring health-related quality of life (HRQOL) using the Rasch model. However, I also note that measures of HRQOL are not theoretically grounded as physical measures are, so there is no theory of the construct to dictate adherence to the strictures of fundamental measurement.
I examine the supposed primacy of physical theory over simple commitments to the fundamental measurement model and find that when physical theory allows more than one model of measurement, we use various epistemic values to choose the measurement model we will use. We tend to choose measurement models with the greatest functional affordances. However, we always do so within the structural constraints imposed by the physical theory.
Social scientists often prefer to use Rasch modeling for their measures because these measures exhibit greater functional affordances than other models – i.e., they purport to be conjointly additive, specifically objective, and to have the potential to be generally objective. However satisfactory fit to a descriptive Rasch model (fit without an appropriate grounding theory) only takes measurement data so far.
Without a theory of the construct to that supports a trade-off relationship between patient ability and item difficulty (and HRQOL measures as a class lack such theory) the construct cannot be guaranteed to be quantitative (i.e., to be conjointly additive), the Rasch model cannot be said to be causal rather than merely descriptive, and, above all, the instrument cannot be said to exhibit fundamental measurement, despite satisfactory fit to a descriptive Rasch model.
Without an appropriate theory of the construct, mere fit to a descriptive Rasch model does not guarantee that we will be able to access all the functional affordances characteristic of fundamental measurement – the affordances advertised by many Rasch proponents. Rasch proponents should be wary of assuming that, merely by fitting data to a descriptive Rasch model, they have bested the functional affordances of ordinal measurement.
University of South Carolina