Several authors recently have discussed the use of interval level statistics with ranked categorical data. Labovitz (1970) and O'Brien (1979) examined the effects via simulation, while Mayer (1971) recommends using robust procedures. Here we show how to score categories in order to facilitate such robust analysis. The RIDIT scoring technique widely used in epidemiology is presented and characterized in this framework, as are conditional mean and median scoring. The relationship between the presented scoring technique, and ordinal ranking procedures are discussed.
Notes
1980–1981 address: Department of Statistics, University of California, Riverside, Riverside, California 92521.