Developing and validating a measuring instrument for the Relationship Harmony personality cluster

Abstract

The object of this study was to develop a valid and reliable measuring instrument for the Relationship Harmony personality cluster as part of the overall SAPI project, a project that aims to develop a valid, reliable, fair and unbiased personality measuring instrument that can be used within the South African context. Due to the large item pool (i.e. 400 items) two questionnaire versions were developed, namely RH–1 and RH–2. A pilot study was conducted with both versions on participating students from tertiary institutions within the North West and Gauteng Provinces (RH–1: n = 507; RH–2: N = 475). Items indicating unacceptable kurtosis were excluded from further analyses due to their unsuitability for factor analysis. Principal component analyses indicated that 31 items from RH–1 and 24 items from RH–2 shared less than 5% of their variance with the total score. These items were thus excluded from further analyses. Principal component analyses were also conducted to determine the correlations between the 23 facets and their relating items. This procedure resulted in another 3 items being removed due to loadings < 0.20. First–order unstructured factor analysis techniques (scree plot, eigenvalues and parallel analysis) indicated that three factors should be retained for RH–1 and two factors for RH–2. Oblique rotations produced factor correlation matrices for both sets of data. Maximum likelihood was used to analyse the factor structure of the Relationship Harmony cluster in both data sets. The data sets were then subjected to higher order factor analysis. A hierarchical Schmid–Leiman factor solution produced a three factor solution for RH–1 (Negative Relational Behaviour, Positive Relational Behaviour and Approachability) and a two factor solution for RH–2 (Positive Relational Behaviour and Negative Relational Behaviour). Construct equivalence across the White and African groups was evaluated by comparing the factor pattern matrices. For RH–1, all facets except for Tolerant loaded on the same factors for both race groups. RH–2’s factors were represented by the same facets irrespective of the race groups. Recommendations were made for future research.