Views on Classical Statistical Modeling in Competition with the Rasch Methodology

Abstract

During 1979–1980 Jan Eric and Peter Allerup implemented a course at the PhD level under the caption: Rasch Models in Social and Behavioral Sciences. It was successfully realized in September/October 1981 near Gothenburg. The late 1970s opened new opportunities regarding the well-known practical problems and limitations of teaching statistical theory supported by empirical analyses on data. In the field of classical factor analyses especially—analyses of high-order contingency tables and analyses carried out by means of Rasch models—such problems were evident. These years provided an efficient means for carrying out such analyses using computers. This chapter will follow some of the tracks used during the course of such research and offers up some reasons why Rasch left classical factor analysis as an analytic statistical method for the benefit of another class of models, originally called Models for Measurement by Rasch himself. A couple of visions will be presented on the factor analysis model Rasch developed in 1953, a few years before the release of his book in 1960. These visions will allow the consideration of various aspects of these Models for Measurement and their extensions, later re-named as Rasch models subsequent to Rasch’s death in 1980.