Date of Award

Spring 1-18-2017

Document Type


Degree Name

Doctor of Philosophy in Industrial/Organizational Psychology (PhD)


Industrial/Organizational Psychology

First Advisor/Committee Member

Dana L. Kendall

Second Advisor/Committee Member

Jerard F. Kehoe

Third Advisor/Committee Member

Lynette H. Bikos


Personality, Factor analysis, Test-retest reliability, Contextualized items


Personality is one of the primary ways that people are distinguished from one another on the basis of their unique tendencies and behavioral patterns. Decades of empirical research have yielded five primary personality traits which have consistently emerged, becoming known as the Five Factor Model (FFM). In particular, the FFM has been widely used in the employee selection realm. However, there have been mixed reviews as to how well the FFM of personality accomplishes that objective, with some research drawing into question the strength of the relationship between personality and job performance.

The purpose of the current investigation is to address these gaps by exploring how a unique personality instrument, developed for commercial applications, may add value for predicting job performance. Archival and anonymous data was collected from the Company for the purpose of investigating the psychometric properties of the proprietary personality selection instrument. The study employed two sets of data, one was a random sample of 10,000 cases for conducting factor analytics and the other was a data set of 1,986 matched cases for investigating reliability. Hypotheses were tested using multiple statistical techniques, such as confirmatory factor analysis (CFA), exploratory factor analysis (EFA), test-retest reliability, Cronbach’s alpha, and Pearson correlations.

Results indicated that the proposed factor structure of the Personality Instrument in which the 15 narrow facets are represented by three broad factors could not be supported (χ2 [3897] = 94548.14, p < .001; GFI = .704; CFI = .433; RMSEA = .048) through CFA. Multiple alternative models were also tested, such as a single-factor model of personality (χ2 [3900] = 96837.26, p < .001; GFI = .694; CFI = .418; RMSEA = .049), the theoretical model poor (χ2 [3910] = 102,525.28, p < .001; GFI = .692; CFI = .383; RMSEA = .050), and also a multitrait multimethod approach (χ2 [3710] = 56336.08, p < .001; GFI = .863; CFI = .671; RMSEA = .038), each of which fell short of showing even moderate model fit to the data. The results of the test-retest reliability analyses showed statistically significant positive relationships between Time 1 and Time 2 Personality Instrument scores (r (905) = .79, p < .01), when administered between 0 and 12 weeks apart, and thus supporting the stability of the Personality Instrument. Compared to test- retest reliability coefficients when taken between 0 and 12 weeks, the test-retest reliability for cases in which the Personality Instrument was taken between 13 weeks and 52 weeks was still statistically significant and positive (r (480) = .67, p < .01), yet a smaller positive relationship as presumed. Test-retest reliability proved to be a better estimate of stability compared to Cronbach’s alpha, which ranged from as low as .05 for Tolerance for Ambiguity to as high as .58 for Affiliative. Lastly, the results investigating as to whether contextualized items within the same substantive facet correlate more so with each other than do generically-worded items.


Copyright Status

Additional Rights Information
Copyright held by author.