"Levels of Measurement" Problem in Communication Research

Abstract

This paper presents an investigation into an important scaling issue: the problem caused by the use of ordinal data in statistical procedures requiring at least interval-level data. Four elements of this problem are presented. First, a sketch of the various approaches to this problem is presented Second the pragmatic and theoretical consequences of violating interval assumptions are outlined. Third evidence is provided which suggests that violations of interval assumptions are widespread in social science research. Finally, a partial solution and two alternative routes to implementation are discussed