Ricardo Primi, Universidade São Fracisco, Brasil & EduLab21, Ayrton Senna Institute, Brasil)
This presentation will discuss the scoring engine of ENEM that uses 3-parameter IRT model when calculating students’ scores. It will show the role of guessing parameter on scoring, its relationship with test unidimensionality and its implications in test score interpretation. It will illustrate these points with data from a sample of 549,265 respondents to the ENEM Mathematical test. It will be shown that assumptions model make about guessing is too strong and inconsistent with empirical data. Presentations will discuss Implications for test score validity and suggest a change in IRT model to mitigate problems encountered with scores interpretation.