Redução da quantidade de itens para estimação de proficiência em teste adaptativo computadorizado com filtro de Kalman

Abstract

Large-scale assessments are important instruments not only in measuring the quality of education, but also in implementing educational policies. However, despite the latest advances, traditional large-scale assessments still face problems, such as the time required for institutions to receive their feedback and the number of items in the test, which is sometimes exhaustive. To cope with this issue, we propose the implementation of a computerized adaptive test based on the item response theory with the application of a Kalman filter to reduce the number of test items. This makes it possible to reduce about 70% of the required number of items in a test, when compared to a traditional test.