Dimensionality of the Creative Personality Scale: Psychometric Network Analysis and a Simulation Study

• In Event: Innovations in Methodological Approaches in Creativity Assessments for Gifted Student Identification

Abstract

Psychometric network analysis (PNA) has been gaining popularity as a promising dimensionality assessment method, and several simulation studies (e.g., Golino & Demetriou, 2017; Golino et al., 2020) have shown that PNA outperforms many traditional methods (e.g., exploratory factor analysis, parallel analysis) in identifying the structure of latent psychological constructs. Some researchers (e.g., Christensen et al., 2019, 2020) have also started to explore the possibility of using the PNA framework to investigate the validity issues of personality measures.
However, PNA has not been widely used yet in the field of creativity assessment due to researchers’ limited understanding of this method and the limited availability of software programs. Additionally, the Creative Personality Scale (CPS; Gough, 1979) has been extensively used to measure creative personality traits. But the factor structure of the CPS has not been fully understood yet. Although previous exploratory factor analysis (EFA) suggested the CPS might contain three factors, it is not clear whether a 4th factor exists (Qian et al., 2019). According to Messick’s (1995) theory of construct validity, score reporting and interpretation (e.g., a total score vs. subscale scores) should reflect the internal structure of a construct. Hence, the purpose of this study is threefold. First, the factor structure of the CPS will be examined using both EFA and PNA, and then confirmatory factor analysis (CFA) will be conducted to test which factor solution is more appropriate, if EFA and PNA produce different factor solutions. Second, the implementation of PNA with R packages (e.g., EGAnet; Golino & Christensen, 2019) will be explored, and step-by-step instructions will be provided so that this study will serve as a tutorial for many researchers who wish to investigate the dimensionality of other psychometric instruments using the PNA framework. Third, although previous simulation studies (e.g., Golino et al., 2020) have investigated the performance of PNA, its effectiveness has not been fully evaluated, especially in the presence of small sample sizes (e.g., N < 500). Therefore, a simulation study will be conducted to further examine the performance of PNA as compared to EFA in the context of the CPS (e.g., binary data) when small sample sizes are present.
Participants consisted of 359 undergraduate and graduate students (232 female), and everyone completed the CPS. EFA and CFA suggested that the CPS might contain four factors (CFI=.943, TLI=.923. RMSEA=.03) instead of three (CFI=.905, TLI=.881, RMSEA=.037), with the four factors being creative, adaptive, individualistic, and conservative. PNA provided evidence for three distinctive communities (i.e., factors)-creative, individualistic, and non-adventurous. But two items (Artificial and Snobbish) appeared to have weak correlations with other items in the non-adventurous community, and further CFA did not support the three-factor solution as an acceptable model fit was not achieved (CFI=.801, TLI=.797, RMSEA=.057). The simulation study is being conducted, and findings will show if PNA performs as well as or even better than EFA in the presence of small sample sizes. Additionally, instructions for using the R packages to conduct PNA will be provided in the full paper.

Authors

• Meihua Qian, Clemson University

• Shenghai Dai, Washington State University - Pullman