ABSTRACT
Theoretical sources on ordinal methods provide formal evaluations of the efficacy of ordinal methods compared to classical ones. Ordinal scales permit much more freedom in the transformation. The idea here was that one could propose that a certain kind of data should fit a particular algebraic model, the algebraic model being one that required interval-level variables. Deriving orders from data rather than purported interval-level variables has considerable justification. The scale distinction that is most salient to the behavioral sciences is that between ordinal and interval scales. Percentiles of that distribution are then used as a basis for inferences about the parameter that is of interest. A familiar application of hypothesis testing is provided by the mean. The concepts of power and robustness apply to confidence interval in a way that parallels their application to hypothesis testing. In hypothesis testing, a null-hypothesis value is established for the parameter on some grounds.
