Evaluation of Chemistry Questions in Concur 2017 Using Item Response Theory

The aim of this study is to evaluate the chemistry questions of university entrance exam (concur) in 2017 by using item-response theory.

This is an applied, descriptive research. 5,000 answer sheets of participants in Empirical Science group, were selected randomly as sample. All parameters were calculated with NOHARM and IRTPRO and EXCEL softwares.

First the initial assumptions of IRT theory (unidimensionality and local Independence) were investigated. The parameters of each questions such as difficulty and discrimination coefficient, were calculated based on the classic and the item response theories. The results showed that 7 questions show a good fit with one and two parameter models and 21 questions are also compatible with three-parameter model and 13 questions don’t fit with any IRT models. The analysis with 3-PL-IRT showed the discrimination coefficient of 26 questions are strong (a> 1.3), 6 are moderate and 3 are weak (a <0.65). Also 11 questions were very difficult (b> 1.2) and 24 questions were appropriate (-1.2

The results showed that 3-PL IRT model has a better fit with the test. All questions are efficient in terms of discrimination coefficient and have a high level of difficulty and also guessing parameter is high in this test. The test’s questions have the highest Information and the least error for a high level of ability (range -1 to 2).