Assessing the factor structure of the South African Personality Inventory by employing a dichotomous and a polytomous response scale

Abstract

In a multicultural society such as South Africa, there is a need for valid and reliable instruments measuring personality. Most personality instruments currently used in South Africa are imported from abroad and therefore have limited utility in the South African context as they have been developed for a specific group. The introduction of technology has resulted in personality measuring instruments increasingly being administered by means of computer-based assessments. The dramatic increase in computer-based assessments has sparked debate regarding the use of various response scale categories in personality assessment.

The present study, which forms part of the broader South African Personality Inventory (SAPI) project, focused on the preliminary qualitative personality structure produced by the SAPI project, which is categorised into nine clusters. The current study aimed to determine whether a dichotomous or a polytomous response scale administered by means of computer-based assessments would be more suitable for measuring the preliminary personality structure of the SAPI. The participants were first- and second-year undergraduate students enrolled at a tertiary institution (N = 490). The inventory consisted of 262 closed-ended personality statements and was administered in both the dichotomous (“agree” and “disagree”) and polytomous (“strongly agree”, “agree”, “somewhat agree/disagree”, “strongly disagree” and “disagree”) response scale form.

The results, which were based on an exploratory factor analysis, revealed that 37.2% of the items in the dichotomous response scale were problematic, whereas only 3.6% of the items in the polytomous response scale were problematic. By comparing the factor structures of the dichotomous and polytomous response scales, the polytomous response scale was determined to be more suitable for measuring the preliminary personality structure of the SAPI.

The conclusion was based on two specific criteria. Firstly, the factor structure across the polytomous response scale loaded similarly to the qualitative personality structure that was conceptualised in the first phase of the SAPI project. Secondly, Cronbach alpha coefficients, ranging from 0.60 to 0.87 across the nine factors, with the exception of the Integrity and Openness clusters with values of 0.45 and 0.53 respectively, for the polytomous response scale were higher than those yielded by the dichotomous response scale.