ASSESSING ATTITUDINAL CONGRUENCE: A CASE FOR ABSOLUTE (AS WELL AS RELATIVE) INDICES

Abstract

Looking through most popular textbooks on research methods and statistics used in the behavioral sciences, one will be unable to find any presentation focused on the measurement of attitude congruence. Yet, the notion of agreement or disagreement between various matched pairs has been widely researched in many areas. Perhaps it has just been assumed that assessing this kind of agreement or congruence is merely an application of well-known methods such as correlational analysis and comparison of means. An example of this situation is found in a recent review of parent-child attitude studies (Sirotnik, 1981) in which three distinct analytic methods seemed to be in use equally often for assessing attitudinal congruence: (a) Pearson product-moment correlation, (b) paired-comparison of means, and (c) comparison of independent group means. Such applications could easily lead to non-convergent interpretations. Moreover, method b is not necessarily sensitive to congruence since the algebraic sign of differences confounds the comparison, and method c is clearly inappropriate given the dependent nature of paired data. It will be argued here that questions of attitude congruence can be of at least two distinct types, absolute and/or relative. These types have attendant differences in analytic models and, therefore, also have different implications for data analysis. The term absolute congruence applies to the simple observed proportion of agreement between respondent pairs. The term relative congruence applies to the family of agreement indices which, although functions of the absolute index, are formulated relative to chance expectations. It should be noted that differentiating between these indices is a general problem pertaining to any form of cross classification for the purpose of measuring agreement. For example, the issue raised herein has recently been confronted and similarly resolved by Kane and Brennan (1980) in the context of criterion-referenced testing and indexing the reliability of mastery-nonmastery decisions.