Abstract
Occupational therapists, like other rehabilitation professionals, have accepted ordinal raw scores as a sufficient basis for developing evaluation tools. This paper summarizes problems commonly found in evaluation methods based on summing ordinal raw item scores and demonstrates how Rasch measurement models provide a solution to the construction of calibrated (linear) measures.
Rasch measurement models are contrasted with Stevens’s lax definition of measurement and Guttman’s unreasonably rigid requirements. The simple Rasch model is a probabilistic formulation of the fundamental requirements for additive linear measurement. This formulation retains Guttman’s concept of order, but construes it probabilistically, making it neither too lax (random) nor too rigid. When a measure is based on a theory of what counts as an observation of more or less of something, Rasch measurement models are useful for constructing valid measures.
