Events and Meetings of Italian Statistical Society, Advances in Latent Variables - Methods, Models and Applications

Font Size: 
The Rasch model for measurement and the role of statistical modelling
Thomas Salzberger

Last modified: 2013-06-16

Abstract


Measurement is an essential constituent of quantitative research. In contrast to the natural sciences, where measurement is a well-defined form of quantification, the social sciences rely on a non-restrictive definition that has given rise to a multitude of alleged measurement models. Against the benchmark of the natural sciences, many of those models raise serious concerns as to their adequacy as measurement models. The Rasch model is a noteworthy exception in this regard. The definition of measurement in the social sciences has also contributed to the fact that measurement is predominantly seen as a statistical task, where the model has to describe given data in the best possible way. By contrast, the psychometric paradigm requires a particular structure in the data that follows from the axiomatic theory of measurement. On the other hand, a comprehensive theory of a given construct that allows for concrete and testable hypotheses is equally important. The psychometric paradigm has to reconcile the requirements of quantity with informative construct theories, while the empirical test of fit of the data to the model, which incorporates those requirements, necessitates statistical modelling. However, in case of a weak construct theory, the application of the Rasch model would at least partly follow the statistical paradigm.

Full Text: PDF