Abstrak Penelitian ini bertujuan untuk mengetahui: (1) karakteristik butir soal Kimia Ujian Nasional berdasarkan teori tes klasik dan teori respon butir; (2) besarnya kecurangan yang terjadi dengan menggunakan Metode Angoff's B-index, Metode Pair1, Metode Pair2, Metode Modified Error Similarity Analysis (MESA) dan Metode G2; (3) metode yang lebih banyak mendeteksi adanya kecurangan dalam pelaksanaan UN Kimia tingkat SMA/MA Negeri tahun pelajaran 2011/2012 di Provinsi Maluku. Hasil analisis

more »

... n pendekatan teori tes klasik menunjukkan 77,5% butir memiliki tingkat kesulitan butir berfungsi baik, 55% butir daya bedanya belum memenuhi syarat, dan 70% butir memiliki pengecoh berfungsi baik dengan indeks reliabilitas tes 0,772. Analisis dengan pendekatan teori respons butir menunjukkan 14 (35%) butir cocok dengan model, fungsi informasi maksimum 11,4069 pada θ =-1,6, dan besarnya kesalahan pengukuran 2,296. Jumlah pasangan yang diduga curang adalah: menurut Metode Angoff's B-index ada 13 pasangan, menurut Metode Pair1 ada 212 pasangan, menurut Metode Pair2 ada 444 pasangan, menurut Metode MESA ada 7 pasangan, dan menurut Metode G2 ada 102 pasangan. Metode yang paling banyak mendeteksi kecurangan secara berturut-turut adalah: Metode Pair2, Metode Pair1, Metode G2, Metode Angoff's B-index, dan Metode MESA. Kata kunci: ujian nasional, karakteristik butir, metode kecurangan Abstract This study aimed to reveal: (1) the characteristics of items of Chemistry Test in National Examination by using the classical test theory and item response theory; (2) the amount of cheating which occured by using Angoff's B-index Method, Pair 1 Method, Pair 2 Method, Modified Error Similarity Analysis (MESA) Method, and G2 Method; (3) the methods that detected more cheating in the implementation of the Chemistry Test in National Examination for high schools in the academic year 2011/2012 in Maluku Province. The results of the analysis with the classical test theory approach show that 77.5% items have item difficulty functioning well, 55% items have discrimination that has not met the requirement yet, and 70% items have distractor that works well with the index reliability test of 0,772. The analysis using the item response theory approach shows that 14 (35%) items fit with the model, the maximum function information is 11,4069 at θ =-1,6, and the magnitude of the error of measurement is 2,296. The number of pairs who are suspected of cheating is as follows: 13 pairs according to Angoff's B-index Method, 212 pairs according to Pair 1 Method, 444 pairs according to Pair 2 Method, 7 pairs according to MESA Method, and 102 pairs according to G2 Method. The most widely detecting cheating in a row is a Pair 2 Method, Pair 1 Method, G2 Method, Angoff's B-index Method, and MESA Method.