Exploring the Effects of Test Length, Number of Alternatives, and Sample Size on the Accuracy of Individual Estimates and Item Parameters within the Framework of Item Response Theory

نوع المستند : المقالة الأصلية

المؤلفون

Assistant Professor, Department of Educational Psychology, Graduate School of Education, Cairo University

المستخلص

the present study aimed to investigate the effects of test length, the number of response alternatives, and different sample sizes on the accuracy of individual ability estimates and item parameters based on Item Response Theory (IRT). The researchers generated sample data using the "WinGen" software, considering various sample sizes ranging from 500 to 2000, different test lengths (ranging from 30 to 50 items), and multiple response formats (three-, four-, and five-option items).
The key findings indicated that the accuracy of ability estimation, as well as the accuracy of the difficulty and discrimination parameters, was not affected by test length. However, the guessing parameter was influenced by test length. Additionally, the accuracy of the theta (θ) parameter was higher with four- and five-option items. In contrast, the discrimination, difficulty, and guessing parameters were not affected by the number of response options. The results also revealed that ability estimation was most accurate with a sample size of 500, whereas the accuracy of the discrimination parameter improved with larger sample sizes, particularly with a sample size of 2000.

ملفات

شارك

إرسال الاستشهاد إلى

الإحصائيات