| The study utilized a unitary approach to evaluate the construct validity of College Basic Subject Examination. The sample consisted of 8009 examinees taking test form LM in year 2000. Psychometric properties of test items and internal consistency of the test were first examined. Second, multiple correlational studies were carried out among subject scores of the test and external criterion. Third, the factorial component of the test was explored based on measured skills. Fourth, the construct structure of the test was tested for each academic subject domain of the test. Fifth, the structural invariance was investigated across different ethnic and gender groups. Finally, differential item functioning analyses were performed on pairs of ethnic and gender groups. The study provided evidence of construct validity for College BASE as an academic achievement test with regard to its item characteristics, reliability, and fidelity to theoretical construct structure. However, the structural noninvariance across ethnic and gender groups indicated some skill scores bear different meanings for different subgroups. In addition, a significant portion of the test items displayed bias against different ethnic and gender groups and contribute to structural noninvariance. The fairness of the test can be improved with revision or deletion of the biased items. Future research may gather predictive-related validity evidence and examine multidimensionality of the test. |
