The Effect of Violating the Unidimensional Assumption on Item and Individual Coefficients and their Suitability to the Four-parameter Model

Abstract

The research aims to investigate the effect of violating the assumption of unidimensionality in the item coefficients (distinction, difficulty, estimation, and lack of interest) and their estimated suitability according to the four-parameter model.

To verify the objectives of the study, the researcher prepared a test consisting of four sub-tests, each consisting of (10) items. Thus, the test as a whole consisted of (40) items of multiple choice type on a sample of (1000) male and female students.

The results showed:

1. With regard to the assumptions of the response theory for the item, the criteria for verifying the unidimensional assumptions, which are the saturation of the paragraphs and internal consistency, are met, while the exploratory factor analysis is not met, and the rest of the assumptions are met.


2. The paragraphs differed in terms of suitability for the model, as the percentage of items that were not suitable for the test in each dimension was (5%), but when ignoring the fulfillment of the unidimensional hypothesis, the number of items that were not appropriate reached (55%)


3. There are statistically significant differences in the discrimination coefficient and its accuracy, the estimation coefficient and its accuracy, and the lack of interest coefficient and its accuracy, according to the fulfillment of the unidimensional hypothesis.


4. There are no statistically significant differences in the difficulty coefficient and its accuracy due to the fulfillment of the unidimensional hypothesis.