Adjustment of the Rasch Multidimensional Model, the Rasch Testlet Model and the Cognitive Diagnosis Model to the 2015 Student Census Evaluation Tests

Abstract

This study aimed at applying the results of three models: a) the Multidimensional Random Coefficients Multinomial Logit Model, b) the Rasch Testlet Model, and c) the Cognitive Diagnosis Model, and comparing those results to the data generated from the math and reading comprehension tests conducted in the 2015 Student Census Evaluation (ECE). The sample consisted of 5,000,000 evaluations of students attending second grade of primary school in educational institutions belonging to the regular basic education program (Minedu, 2009). The results showed that: 1. there is a high correlation between modeled dimensions in the multidimensional model both in the math and reading comprehension tests, 2. it is convenient to apply a one-dimensional model to the data derived from the application of the ECE tests, and 3. there is some consistency between the Rasch Testlet Model and the Cognitive Diagnosis Model.