The revision and analysis of parallel forms of a test of human relations insight

The main purpose of this study was to revise Supervisors Problems Forms XA and XB for Improved reliability. Additional objectives included an item analysis, a factor analysis, and a validation study of the revised forms. Forms XA end XB each contained eight cases and fifty questions. K-R 20 reliability estimates were 0.64 and 0.66 respectively, and equivalent form reliability was 0.49, all based on college student data. The revision involved the use of fewer questions per case plus the addition of new cases and questions. The revisions, labeled Forms XL and XM, were administered to a sample of supervisors (N=118) at a chemical plant and to a sample of college students (N=120). Separate analyses were performed for the industrial and student groups on each test form. Results of the item analysis of Forms XL and XM are presented in Table V. The distribution of test scores showed no significant departure from normality. A comparison of test means and standard deviations indicated that Forms XL and XM satisfied these criteria for parallel tests. K-R 20 reliability estimates were 0.76 on both forms for the supervisors and for the students 0.62 and. 0.70 for Forms XL end XM respectively. Equivalent form reliability was 0.83 for a "technical" subgroup of supervisors (N=37), 0.87 for a "nontechnical" subgroup (N=81); and 0.73 for the student sample. The method of intraclass correlation was used to estimate the reliability of performance ratings for the supervisors. Test scores were correlated with these ratings and with selected personal data. Correlations generally were significant and can be found in Tables V and VI. The test forms also discriminated between a group of shift foremen (N=11) seected on the basis of merit and a second group (N=22) selected on the basis of seniority. In view of correlations ranging as high as 0.59 between Supervisors Problems and the Wonderlic Personnel Test, semipartial correlations were computed between Supervisors Problems and the various rating traits to estimate the relationships with the effect of intelligence removed from the test scores. While substantial increases or decreases in correlation were observed for several traits, the results did not reflect a consistent pattern. Test scores for both the student and supervisory groups were factor analyzed by means of a principal factors solution with a varimax rotation of factors. Resulting factors proved to be uninterpretable.