Tri aplicada a análise de itens para o ensino de Estatística

Abstract

The purpose of this research is the development and storage of a collection of selected items that aims to design assessment tools, which may help the educator of the Basic Statistics subject. Therefore, concepts and fundamentals of the Item Response Theory (IRT) were used in order to prepare instruments that contribute to an assessment of the teaching-learning process. The architecture of a collection of items calibrated by IRT depends on the model parameter adopted. On this thesis, the three parameter logistic unidimensional model will be adopted, which emphasizes the characteristics inherent to the individual, his or her latent trait and the characteristics: the discrimination (a), dificulty (b) and pseudo-guessing (c) parameters and the item characteristic curves (ICC). The same curve was used in two different classes to build the collection of items. This collection of items was obtained through a test applied to the students from all courses of the Federal University of Lavras (UFLA) taking the leveling teaching subjects , which is offered by the Department of Exact Sciences. The test was taken by a total of 50 students, divided into two groups, one with 24 students and another with 26. The analyses were performed using the open source statistical software called IRT Command Language.