Abstract
The Positive Social-Relational factor of the South African Personality Inventory (SAPI) addresses the social collectivistic nature of the diverse South African population. The study investigated to what extent two indigenous language versions of the SAPI have different probabilities of endorsing personality items, compared to the original source language. The presence of differential item functioning (DIF) was therefore investigated for the English (SAPI-EN; n = 3,912), the Southern Sotho (SAPI-SS; n = 293) and Tshivenda (SAPI-VE; n = 290) language versions of the SAPI. This quantitative study made use of secondary data that employed the purposive sampling strategy. To meet the objectives of the study, the Rasch Unidimensional Measurement Model (RUMM2030) was used. The data from the three language versions showed unidimensionality across the six facets of the Positive Social-Relational factor. However, the Social Intelligence facet indicated an inadequate fit to the Rasch model; therefore, no further analysis could be conducted. The remaining five facets expressed uniform DIF, with items revealing more discrimination for the Sesotho language version. The overall fit to the Rasch model was improved after items with DIF were resolved for the five facets. The study aims to contribute to the continuous development of the SAPI and to further improve on the application of personality assessments in multilingual and cross-cultural workplace settings such as the South African context. Keywords: South African Personality Inventory (SAPI), Differential item functioning (DIF), Positive Social-Relational, Rasch analysis, Simultaneous test development.
M.Com. (Industrial Policy)