The application of Rasch techniques to provide measurement and obtain meaning from cross-tabulated data produced by standard statistical surveys is described by Otis Dudley Duncan in "Surveying Subjective Phenomena" (Turner & Martin, NY: Russell Sage, 1984)
In Vol. 1, Section 6.4, Duncan reviews the measurement properties of the Rasch model and applies it to a cross- classification of responses to questions in the 1977 General Social survey. He uses the Rasch model to determine whether local independence between classification cells is approximated and to identify significant interaction. Once the rows or columns showing significant non-independence are eliminated, the ratios between row margins become the odds estimates of the social entities (demographics) and similarly the ratios of the column margins become the odds estimates of the dispositions (attitudes). These estimates define the variable. In Vol. 2, Section 12, Duncan extends this to other types of attitude surveys in which responses are classified and counted by group.
Duncan concludes that "As long as we are content merely to find models sufficiently flexible to fit available [social survey] data acceptably well, only by happenstance will we achieve what can be termed measurement.. Nor will we know whether Rasch measurement can be carried out in a deliberate way in surveys until we try very hard to do it.. Nevertheless, as the example from the Elmira panel study [of a political campaign] suggests, under favorable conditions we may actually be able to tell the difference between changes in disposition and changes in the social entity to which the disposition relates, thereby achieving the separation of the measures of the two things that only Rasch's approach provides" (p.400).
Survey Research, O D Duncan … Rasch Measurement Transactions, 1991, 5:2 p. 143
The URL of this page is www.rasch.org/rmt/rmt52e.htm
Website: www.rasch.org/rmt/contents.htm