120 likes | 277 Views
Assessment of Learning: Test Design and Administration Factors That Affect Student Performance. Rita Czaja and Scottie Barty Northern Kentucky University Highland Heights, KY. Research Motivation. New AACSB accreditation standards Unexpected results on department’s self-assessment test.
E N D
Assessment of Learning: Test Design and Administration Factors That Affect Student Performance Rita Czaja and Scottie Barty Northern Kentucky University Highland Heights, KY
Research Motivation • New AACSB accreditation standards • Unexpected results on department’s self-assessment test
Interpreting Percentile Ranks • Percentile ranks • May be better than the percent correct for comparisons across test sections • If some sections have more difficult questions • Performance is imprecise measure of difficulty • Reflects design of test questions and students’ abilities
Evaluation of Test Questions • Rule for effective multiple-choice questions • Scored the design of each question: • 1 = Simple (Straightforward computation; basic terms; single topic) • 2 = Moderate (One less familiar term; two-step logic process; some extraneous data) • 3 = Complex (Problem set-up not obvious; unusual situation; complex formula, definition, or rule; multiple topics; much extraneous data)
Hypotheses H1: Seniors’ average percent correct is negatively related to the difficulty of a question’s design. H2: Seniors’ average percent correct is positively related to the extent of repetition. H3: Seniors’ average percent correct is negatively related to the difficulty of recalling a question’s content.
Descriptive Statistics • Question-Design • Financial: 2.37 average • Managerial: 1.91 average • Repetition (number and percent of questions covered in principles and intermediate) • Financial: 15 (39%) • Managerial: 16 (70%) • Ease-of-Recall • Financial: 2.03 average • Managerial: 1.74 average
Table 3: Regression ResultsSenior’s Percent Correct = a + b1 Design + b2 Repetition + b3 Ease of Recall Residual + b4 Test Section
Lessons Learned • Be very clear about what you want to measure. This decision drives • the timing of testing • the nature of questions used • the level of performance that can reasonably be expected