UJIAN MELALUI GFORM, MUDAHKAH DITEBAK? ANALISIS ITEM RESPONSE THEORY 3 PARAMETER LOGISTIK
Quality assessment is essential for measuring students' abilities, but challenges such as guessing in multiple-choice questions can reduce test validity. This study aims to evaluate the quality of test items using the 3 Parameter Logistic (IRT 3PL) Item Response Theory model. A total of 191 students answered 40 multiple-choice questions. The analysis included item-total correlation, point biserial correlation, and IRT 3PL parameters (discrimination, difficulty, guessing). Results showed that most items had good quality (A 1; p 0.05), although some items, such as item 6 and 31, required revision. Conclusion: The test instrument is adequate, but items with high guessing probabilities or low discrimination need improvement to enhance the test's validity and reliability. Penilaian berkualitas penting untuk mengukur kemampuan peserta didik, tetapi tantangan seperti menebak dalam soal pilihan ganda dapat mengurangi validitas tes. Penelitian ini bertujuan mengevaluasi kualitas butir soal menggunakan Model Item Response Theory 3 Parameter Logistic (IRT 3PL). Sebanyak 191 mahasiswa mengerjakan 40 soal pilihan ganda. Analisis meliputi korelasi item-total, point biserial, dan parameter IRT 3PL (daya pembeda, tingkat kesulitan, peluang menebak). Hasilnya sebagian besar butir memiliki kualitas baik (A 1; p 0,05), meskipun beberapa butir seperti item 6 dan 31 memerlukan revisi. Kesimpulan: Instrumen tes memadai, tetapi butir dengan peluang menebak tinggi atau daya pembeda rendah perlu diperbaiki untuk meningkatkan validitas dan reliabilitas tes.
Authors
- Soeprijanto Soeprijanto (ORCID: https://orcid.org/0000-0001-9447-698X)
- Helli Ihsan (ORCID: https://orcid.org/0000-0002-9959-8609)
- Awaluddin Tjalla (ORCID: https://orcid.org/0000-0001-6204-9932)
Publication Details
- Journal
- EDUTECH
- Published
- 2025-04-16
- DOI
- https://doi.org/10.17509/e.v24i1.77831
- Primary Topic
- Management and Optimization Techniques
- Type
- article
- Field-Weighted Citation Impact
- 0.00