Programmeerimise nädalatestide vastuste analüüsija

Abstract

In the bachelor’s course Programming (LTAT.03.001) at the University of Tartu, weekly Moodle-based tests are used to support learning. The success of such tests depends on the quality of individual questions, but current analysis tools are often insufficient to assess it. This thesis presents a Python-based tool that analyzes test question quality using statistical indicators from classical test theory and item response theory. The tool also compares each student's first and best attempt to assess whether a question promotes learning. Testing showed that a large portion of questions were too easy or had low discrimination, failing to fulfill their purpose effectively. The developed application supports teachers in identifying weak questions and improving overall test quality.