Avaliação em Larga Escala como Política do Estado: um estudo comparativo entre a teoria clássica dos testes e a teoria da resposta ao item na Olimpíada Brasileira de Matemática das Escolas Públicas (OBMEP).

Abstract

The Brazilian Olympiad of Mathematics of Public Schools - OBMEP has been held in Brazil since 2005, and uses as a criterion / model for the correction of the tests of its first phase the Classical Theory of Tests - TCT, model in which the result of the test is obtained computing the quantity of items correctly answered by each participating student. An alternative correction proposal for OBMEP is presented in this research, the Item Response Theory (TRI), which considers information contained in the evaluated items, such as the degree of difficulty and even the possibility of random responses by the respondents, without having sufficient skills to correctly answer certain items, the so-called "kick", thus arising, the need to understand the possible contributions of TRI in the process of large-scale evaluation of students submitted to OBMEP. In this sense, the objective of the research is to discuss the large-scale evaluation in Brazil and to measure the proficiency in mathematics of students submitted to level 1 OBMEP of public schools in the city of Santarém-Pa. For the theoretical reference, we list and analyze authors who study large-scale evaluations (PASQUALI and PRIMI, 2012, Tavares, 2000, Werle, 2010, Quartes, 2014 and Andrade 2000 and VALLE 2000) as well as authors (AFONSO, 2009 ; FREITAS, 2007), who discuss the reform of the State and the consolidation of the evaluating State that establishes the external evaluation policy for basic education. The results showed that the test presents high di-culty index and low internal consistency, which made difficult the analysis via logistic model of 3 parameters of TRI. Thus, a pedagogical analysis of the items of the test was necessary contributing to a better interpretation of the performance of the students submitted to OBMEP.