Validity, reliability and equivalence of parallel examinations in a University setting
Malau-Aduli, BS and Walls, J and Zimitat, C, Validity, reliability and equivalence of parallel examinations in a University setting, Creative Education, 3, (6A) pp. 923-930. ISSN 2151-4755 (2012) [Refereed Article]
A key issue to address in the design and implementation of any assessment system is ensuring its reliability and validity. University assessment policies often require staff to prepare parallel examinations for students who are unable to sit the initial examination. There is little published literature to give confidence to staff or students that these examinations are indeed reliable or equivalent. This study was conducted to determine the validity, reliability and equivalence of two parallel examinations that have been developed under highly defined quality assurance (QA) processes in a university setting. Collated assessment results for all the 76 participants who sat the parallel examinations were subjected to statistical and correlational analysis to test for significant differences between mean scores and their associated standard deviations. Item analysis was conducted for each assessment by computing the difficulty index (DIF), discrimination index (DI) and Kuder-Richardson 20 (KR-20) reliability using classical test theory. Results indicated comparative proportions of difficulty, functional distractors and internal consistency of the assessment items on both examinations. Comparison of student performances in both examinations revealed that there was no significant difference in mean scores. However, a highly positive and significant correlation (r = 0.82) between student total scores in both examinations was evident. Approximately two thirds (62.5 %) of students with low scores in the first examination also achieved low scores in the second examination. Furthermore, two thirds of the students were ranked in the same order based on performance in both examinations. The established QA processes for assessment in the school provided a strong basis for the generation of multiple sources of data to support arguments for the validity of examinations. It is possible to develop valid, reliable and equivalent parallel tests in university settings with the presence of well-defined QA processes.