Article Per Year (5 Year)
p-Index From 2021 - 2026
0.61
P-Index
This Author published in this journals
All Journal Jurnal Penelitian dan Evaluasi Pendidikan REiD (Research and Evaluation in Education)
Ibnu Rafi
National Sun Yat-sen University
Decision Sciences, Operations Research & Management Education

Published : 3 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 3 Documents
Search

 1 
Equating high-stakes junior high school mathematics tests using various methods: Which method is the most advantageous? Sabarudin Sabarudin; Sri Yanto; Erna Andriyanti; Heri Retnawati; Ezi Apino; Ibnu Rafi; Sefina Triastuti
REID (Research and Evaluation in Education) Vol. 11 No. 2 (2025)
Publisher : Graduate School of Universitas Negeri Yogyakarta & Himpunan Evaluasi Pendidikan Indonesia (HEPI)

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.21831/reid.v11i2.95819

Abstract

A significant concern about test instruments comprising various test forms is the equivalency among those forms. When the test instrument is deemed "high-stakes," with results commonly utilized to ascertain students' passing status, the equivalence between test forms must be a paramount consideration. The goal of this study is to explain and compare the four approaches used to equate test forms in high-stakes arithmetic tests: mean-mean, mean-sigma, Haebara, and Stocking-Lord. The test used was the national mathematics exam for junior high school students, which had five sets of questions. There were 33,451 students who took the national test for junior high school in Indonesia and took part in the study. An equivalent groups design was used to establish equivalence. We used four equating methods: mean-mean, mean-sigma, Haebara, and Stocking-Lord, to do equating based on item response theory. All analyses were performed via R Studio. The study results indicated that the five test types employed in high-stakes mathematics assessments were quite comparable, with the Stocking-Lord technique yielding the most equal scores in comparison to alternative methods. We spoke about all the results and possible areas for future research.
Optimal number of response categories for a Likert-type questionnaire: A simulation study within the frameworks of Classical Test and Item Response Theory Neng Ida Soniawati; Ezi Apino; Heri Retnawati; Farida Agus Setiawati; Ibnu Rafi; Munaya Nikma Rosyada; Rezkilaturahmi Rezkilaturahmi
REID (Research and Evaluation in Education) Vol. 12 No. 1 (2026)
Publisher : Graduate School of Universitas Negeri Yogyakarta & Himpunan Evaluasi Pendidikan Indonesia (HEPI)

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.21831/reid.v12i1.95846

Abstract

Likert-type questionnaires have been widely used in educational contexts with many types of response categories. The use of many response categories raises the question of the optimal number of response categories for a Likert-type questionnaire. This issue continues to be investigated by researchers, and the current findings present an inconsistent answer. This study aims to find the optimal number of response categories in developing the Likert-type questionnaire based on classical test theory (CTT) and item response theory (IRT) framework. The simulation design used 8 (number of response categories: 3- to 10-point scale) × 3 (number of items: 20, 25, and 30 items) × 5 (sample size: 100, 200, 500, 1000, and 5000) = 120 conditions. We generated a dataset for each condition for 1000 replications using WinGen 3 and randomly took 10 data points from each dataset to estimate its psychometric properties using the CTT and IRT analysis frameworks. The optimal number of response categories was determined based on the highest reliability coefficient, total explained variance, and information function values. Using the CTT analysis framework, the simulation results indicated that the 8-point scale is optimal for Likert-type questionnaires with 20 and 30 items and for all sample size variations. However, for questionnaires with 25 items, the 8-point scale is only optimal for sample sizes of 100 and 5000, while for sample sizes of 200 to 1000, the 10-point scale is more optimal. Using the IRT analysis framework, the simulations reveal that it is necessary to consider the number of items and the sample size to determine the optimal number of response categories. In this context, 8- and 10-point scales are generally more recommended for Likert-type questionnaires. The findings provide researchers with insight into the optimal number of response categories and the analytical model for developing a Likert-type questionnaire.
Construct validity and reliability of a high-stakes mathematics test for junior high school: A case from the Indonesian national examination Astutiati Astutiati; Muhammad Kusnun Muhsin; Raden Rosnawati; Heri Retnawati; Ezi Apino; Ibnu Rafi; Umi Farisiyah
Jurnal Penelitian dan Evaluasi Pendidikan Vol. 30 No. 1 (2026)
Publisher : Graduate School, Universitas Negeri Yogyakarta in cooperation with Himpunan Evaluasi Pendidikan Indonesia (HEPI) Yogyakarta

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.21831/pep.v30i1.95822

Abstract

High-stakes tests are used to inform critical decisions concerning students, teachers, schools, and educational policies. Therefore, they must meet rigorous quality-assurance standards. One key approach to ensuring the quality of high-stakes tests is establishing construct validity and reliability. This study aims to prove the construct validity and estimate the construct reliability for a high-stakes mathematics test for junior high school students. This study uses response data from participants taking the national junior high school mathematics exam in Indonesia. A simple random sampling technique was used to select 300 samples. Data were analyzed using second-order confirmatory factor analysis with the help of the R program. The study results revealed four factors that make up the mathematical ability construct: number, algebra, geometry, and statistics. All manifest variables (18 indicators) confirmed a significant contribution. The model fit indicators are met (chi-square = 140.013, df = 131, p = 0.279, RMSEA = 0.015, CFI = 0.994, TLI = 0.993; SRMR = 0.039), so the construct measured in the high-stakes mathematics test is valid for measuring students’ mathematical abilities. Thus, the test construct has a reliability coefficient of 0.98, reflecting that the construct is reliable for measuring mathematical abilities.
Co-Authors Astutiati Astutiati Erna Andriyanti Ezi Apino Farida Agus Setiawati Heri Retnawati Muhammad Kusnun Muhsin Munaya Nikma Rosyada Neng Ida Soniawati Raden Rosnawati Rezkilaturahmi Rezkilaturahmi Sabarudin Sabarudin Sefina Triastuti Sri Yanto Umi Farisiyah