ABSTRACT
This chapter deals with very weak tests involving ordinal measures of association that can be used to make inferences concerning underlying frequency distributions, nonlinearity, nonadditivity, or homoscedasticity and that may make it possible to transform the ordinal data into crude interval scales. It argues that many of the variables treated as ordinal scales can be conceived at somewhat better than an ordinal level, perhaps that of an ordered-metric scale for which distances between points can be ordered. The chapter discusses briefly what appear to be some of the most important kinds of variables often treated at the ordinal level. In many instances indirect measurement and inadequate conceptualization are confounded, with the result that the investigator selects an ordinal measure because he wishes to test a theoretical proposition on a higher level of abstraction than that directly implied by the indicators.
