Improving assessment quality: Development of evaluation instruments for sixth-grade Mathematics learnin

Valid and reliable assessment instruments are crucial for accurately measuring student competencies, yet many fail. This study aims to develop a Mathematics learning outcome assessment instrument based on the Kurikulum Merdeka for sixth-grade elementary school students, focusing on validity and reliability using the Rasch model. This instrument is designed to encompass cognitive, affective, and psychomotor competencies following the needs of the Kurikulum Merdeka. Analysis was conducted on 213 students using a quantitative approach with Winsteps software to evaluate item quality, unidimensionality, reliability, difficulty level, and potential differential item functioning (DIF). The research results indicate that the instrument is highly reliable and meets the unidimensionality criteria. The distribution of question difficulty levels varies from very easy to very difficult, reflecting the instrument's ability to measure students with a range of abilities. All items meet the fit criteria based on fit statistics (Outfit MNSQ, ZSTD, and Point Measure Correlation). However, two items (Item 1 and Item 5) show significant DIF bias based on gender analysis. This study concludes that this assessment instrument based on the Kurikulum Merdeka is valid, reliable, and suitable for assessing students' mathematical abilities. Abstrak Instrumen penilaian yang valid dan reliabel sangat penting untuk mengukur kemampuan peserta didik secara objektif, namun banyak instrumen yang ada belum merepresentasikan kompetensi peserta didik dengan akurat. Penelitian ini bertujuan untuk mengembangkan instrumen penilaian hasil belajar matematika berbasis Kurikulum Merdeka untuk peserta didik kelas VI Sekolah Dasar, dengan fokus pada validitas dan reliabilitas menggunakan model Rasch. Analisis dilakukan pada 213 peserta didik menggunakan pendekatan kuantitatif dengan software Winsteps untuk mengevaluasi kualitas item, unidimensionalitas, reliabilitas, tingkat kesulitan, dan potensi bias diferensial (Differential Item Functioning/ DIF). Hasil penelitian menunjukkan bahwa instrumen memiliki reliabilitas tinggi dan memenuhi kriteria unidimensionalitas. Sebaran tingkat kesulitan soal cukup bervariasi dari sangat mudah hingga sangat sulit, mencerminkan kemampuan instrumen dalam mengukur peserta didik dengan beragam tingkat kemampuan. Semua item memenuhi kriteria kesesuaian berdasarkan statistik kecocokan (Outfit MNSQ, ZSTD, dan Point Measure Correlation), namun terdapat dua item (Item 1 dan Item 5) yang menunjukkan bias DIF signifikan berdasarkan analisis gender. Kesimpulan dari penelitian ini adalah bahwa instrumen penilaian berbasis Kurikulum Merdeka ini valid, reliabel, dan cocok untuk digunakan dalam menilai kemampuan matematika peserta didik. Kata Kunci: evaluasi instrumen; Kurikulum Merdeka; Model Rasch; penilaian pendidikan; reliabilitas; validitas