An Evaluation of the Psychometric Properties of the STPI (Form Y) for South African Students

Abstract

This study investigated the psychometric properties of the State Trait Personality Inventory (STPI-Y) with a South African student sample. Both classical test theory and item response theory (Rasch Model) were used. The sample comprised 2298 students. The students ranged in age from 17 years to 51 years. The reliability of the STPI-Y was evaluated using both the Cronbach’s alpha internal consistency reliability and the person and item separation indices obtained using Rasch analysis. The Cronbach’s alpha reliability coefficient for all eight scales ranged from satisfactory (.72) to excellent (.92). The person separation indices were mostly below .80, which indicates that more items need to be added to these scales to adequately distinguish between those with higher levels and those with lower levels of each state or trait. The item separation indices were generally above .90, indicating that the sample was large enough to confirm the hierarchy of item difficulties. Rasch analysis of the fit of the different scales of the STPI-Y indicated that two items on the State Curiosity scale and one item on the Trait Curiosity scale appear to measure a different dimension from the remaining items on each scale. Exploratory and confirmatory factor analysis also indicated that certain items on the State Curiosity and Trait Curiosity scales had poor factor loadings. These findings are consistent with previous research studies. Nonparametric statistics were used to determine if differences existed between gender and ethnic groups. With the exception of the Trait Anger scale, statistically significant median differences between the gender groups were reported on each of the scales of the STPI-Y. Statistically significant differences were also found when comparing ethnic group medians. The effect sizes of these differences, however, were below .10 and are unlikely to impact the interpretation of the scores across the groups. Uniform differential item functioning (DIF) was used to explore possible bias between gender and ethnic groups. There was minimal evidence of DIF when comparing gender groups. When comparing ethnic groups, there was little evidence of DIF on the State and Trait Curiosity scales. Differences on the remaining scales were across different ethnic groups and in different directions, therefore, it was concluded that these results are unlikely to result in bias at scale level. The suitability of the language of the STPI-Y items was also investigated. Some items were found to be problematic. However, if these items were deleted, little change would occur in the reliability of their scales. It appears that the STPI-Y scales have acceptable psychometric properties for the South African student population. However, the State and Trait Curiosity subscales need to be interpreted with caution. It is recommended that the tool is constantly refined and researched to continually improve the quality of the instrument.