On scoring ordered classifications
Antony Fielding · British Journal of Mathematical and Statistical Psychology · 1997
Methods for scoring ordered classifications in a variety of methodological situations are reviewed. These contrast with model‐based approaches through such foci as generalized linear models. The practical empirical advantages of scoring functions are discussed. In developing and extending previous work on axiomatic frameworks for scoring functions it is suggested that conditional mean scoring functions have a logical and intuitive appeal. Extensions to axiom‐based criteria are suggested and their form is suggested by a convexity theorem. This is extended to other useful properties. The debate surrounding choices of arbitrary scores is discussed and some application areas are reviewed. It is concluded that some minimal attention be given to forms of scores whilst recognizing that scoring will continue in practice and that controversy will continue.