RESULTS OF PSYCHOMETRIC ANALYSIS OF TEST ITEMS IN THE KAZTEST SYSTEM

Abstract

This article presents a comprehensive psychometric analysis of the listening and reading sections of the KAZTEST system based on the Rasch model using the Winsteps 5.8.3.0 software. This approach ensures objective evaluation of task quality, test balance, and test-takers’ language abilities. The analysis examined test item difficulty, reliability, separation indices, and test-taker performance indicators. The results confirmed the high quality of KAZTEST test items: Item Reliability – 0.99, Person Reliability – 0.76, and Cronbach’s Alpha – 0.84, demonstrating strong internal consistency and psychometric stability. Data from the listening section showed a clear differentiation of comprehension levels among test-takers. However, to enhance the accuracy of low-level performance assessment, it is recommended to allow two listenings of the audio texts, a practice widely used in international tests such as TOEFL, IELTS, and TestDaF. In the reading section, item difficulty ranged from –5.12 to +2.63 logits, confirming the test’s ability to reliably measure language proficiency across basic, intermediate, and advanced levels. The balance of text content, length, and linguistic complexity supports test validity and content representativeness. Overall, the findings provide a solid empirical basis for improving the psychometric quality of KAZTEST and enhancing its alignment with international language testing standards. Recommendations include optimizing the structure by difficulty levels, increasing cognitively demanding tasks, and revising test specifications according to Rasch model data. The study contributes to strengthening the reliability, fairness, and interpretive accuracy of the national language testing system.