Multiple R, the “Parametric Strategy,” and Measurement Imprecision

Abstract

This analysis examines the efficacy of using “the parametric strategy” when doing multiple regression analyses using multiple R and related statistics. The parametric strategy involves transforming ordinal scale data into interval scale measures by assigning rank ordered equal interval scale values to ordinal categories. Studies that have examined the adequacy of the parametric strategy in regression analyses have (1) examined only bivariate, and not multivariate, relations, (2) usually not adequately examined measurement imprecision effects due to ordered polytomies in the data and (3) used simulation data, not real data from substantive sociological contexts. In this study an experiment using real data is undertaken to assess measurement imprecision effects in multiple regressions; observed values of multivariate regression statistics are found to diverge substantially from true values where the parametric strategy is applied to data involving ordered polytomies. The implications of these findings are discussed.