摘要：Penelitian ini bertujuan untuk (1) mengidentifikasi karakteristik butir-butir tes pada perangkat soal ujian nasional mata pelajaran Matematika tingkat SMP tahun pelajaran 2009/2010 yang dikalibrasi dengan metode kalibrasi fixed parameter , dan (2) mengetahui metode kalibrasi fixed parameter yang paling akurat di antara metode NWU-OEM ( no prior weights updating and one expectation-maximization cycle ), NWU-MEM ( no prior weights updating and multiple expectation-maximization cycles ), OWU-OEM ( one prior weights updating and one expectation-maximization cycle ), OWU-MEM ( one prior weights updating and multiple expectation-maximization cycles ), dan MWU-MEM ( multiple weights updating and multiple expectation-maximization cycles ). Penelitian ini menggunakan pendekatan kuantitatif deskriptif. Subjek penelitian adalah data respons ujian nasional mata pelajaran Matematika tingkat SMP tahun pelajaran 2009/2010 dari provinsi DI Yogyakarta. Kriteria akurasi metode adalah nilai fungsi informasi tes dan kesalahan pengukuran. Hasil penelitian adalah sebagai berikut. (1) Statistik parameter butir-butir tes pada perangkat ujian nasional mata pelajaran Matematika tingkat SMP tahun pelajaran 2009/2010 menunjukkan rerata indeks daya beda butir berada pada interval [1,07 sampai 1,14], rerata indeks kesukaran butir [-0,35 sampai -0,20], dan rerata pseudo guessing < 0,25. Nilai theta-nilai kemampuan-pada posisi fungsi informasi butir menjadi maksimal menunjukkan grafik fungsi kelima metode kalibrasi fixed-parameter hampir berimpit. (2) Metode OWU-OEM merupakan metode yang paling akurat dalam mengestimasi parameter butir pada perangkat tes ujian nasional mata pelajaran Matematika tahun pelajaran 2009/2010. Kata kunci : akurasi, kalibrasi, fixed parameter, algoritma, Expectation-Maximization ______________________________________________________________ THE ACCURACY OF THE FIXED PARAMETER CALIBRATION METHOD:STUDY OF MATHEMATICS NATIONAL EXAMINATION TEST Abstract This study aimed to: (1) identify the characteristics of the test items on the mathematics test of the national examination which are calibrated with the fixed parameter calibration methods, and (2) reveal the most accurate fixed parameter calibration methods among NWU-OEM (no prior weights updating and one expectation-maximization cycle), NWU-MEM (no prior weights updating and multiple expectation-maximization cycles), OWU-OEM (one prior weights updating and one expectation-maximization cycle), OWU-MEM (one prior weights updating and multiple expectation-maximization cycles), and MWU-MEM (multiple weights updating and multiple expectation-maximization cycles) methods. This study used descriptive quantitative approach. The subject is the testee’ responses to the mathematics national examination in junior high school in 2009/2010. The criteria of the accuracy methods are TIF and SEM. The research results are as follows. (1) Item of statistical parameter on Mathematics national examination test in 2009/2010 showed the average of item discrimination on the interval [1.07, 1.14], the average of item difficulty on the interval [-0.35, -0.20], and the average of pseudo guessing is c < 0.25. Theta - ability - score where the item information function maximalist showed the function of five fixed-parameter calibration methods almost coincides. (2) OEM-OWU method is the most accurate in estimating the parameters on mathematics national examination test in 2009/2010. Keywords : Accuracy, Calibration, Fixed Parameter, Algorithm, Expectation-Maximization
