Students' Performance and Tests' Discriminating Power: Experiments Involving a Database Management System Course

摘要

Assessment represents a relevant branch of educational research, mainly because evaluating and measuring students' learning experiences outcomes are not simple tasks. This paper aims at investigating students' overall performance and the discriminating power of particular tests' items in the context of business education (accountancy). The purpose of this paper is to contribute with this issue while analyzing it, with scientific approach, from an accounting information systems standpoint: two experiments based on a database management system (DBMS) undergraduate course.rnThe discriminant analysis generated discriminant functions with high canonical correlations (Experiment 1 = 0.898 and Experiment 2 = 0.789). As a result, high percentages of original grouped cases were correctly classified (Experiment 1 = 98.5% and Experiment 2 = 95.2%). Even considering the existence of significant correlations between questions (items) and grade (performance), the discriminant analysis presented a relatively small number of them with high discriminant conditions: 31.8% of the items within Experiment 1 (multiple-choice), and 50% of them within Experiment 2 (short-answer).rnAccording to these findings, especially in a business education context, instructors and institutions should analyze and try to improve their assessment methods and techniques, mainly in order to be of no influence (or minimum influence) regarding evaluating students' performance.