Penelitian ini bertujuan untuk mengetahui (1) akurasi metode linear dan ekipersentil pada pendekatan Klasik dan mean mean dan mean sigma pada pendekatan teori respons butir; (2) karakteristik tes prestasi belajar matematika model campuran kelas III, IV, V, dan VI Sekolah Dasar yang digunakan dalam penelitian ini; dan (3) perkembangan kemampuan siswa (θ) kelas III, IV, V, dan VI pada pelajaran matematika berdasarkan tes yang digunakan. Data penelitian adalah skor siswa pada ujian akhir semester tahun 2013 yang dikumpulkan dengan desain tes dengan butir bersama. Hasil penelitian menunjukkan bahwa, (1) Standart Error of Equating (SEE) metode ekipersentil 0,505116 dan metode linear 0,608, sedangkan Root Mean Square Different (RMSD) metode mean sigma 0,344 dan metode mean mean 0,394, perbedaan nilai SEE pada pendekatan Klasik dan RMSD pada pendekatan IRT sangat kecil sehingga kebermaknaannya masih perlu diteliti; (2) tes prestasi belajar matematika kelas III, IV, V, dan VI berturut-turut memiliki rerata location katagori sedang -0,070, -0,228, -0,324, -0,430, slope yang baik 0,537, 0,558, 0,639, 0,644, dan items fit statistics minimum 0,077, 0,052, 0,055, dan 0,198; (3) keempat metode penyetaraan memberikan informasi bahwa nilai θ meningkat seiring dengan meningkatnya peringkat kelas, dengan rerata θ kelas III, IV, V, dan VI berturut-turut sebesar -0,104, -0,076, 0,004, dan 0,173. Namun, kenaikan nilai tersebut sangat kecil sehingga kebermaknaannya masih perlu diteliti.Kata kunci: Metode penyetaraan, Tes model campuran ______________________________________________________________ VERTICAL EQUATING USING CLASICAL METHOD AND ITEM RESPONSE THEORY IN ELEMENTARY SCHOOL STUDENTSAbstract The study was aimed at finding: (1) the accurate linear method and equipercentile method in Classic approach and meanmean and mean sigma method in IRT approach, (2) the characteristics of the mathematic mixed-model achievement test for elementary students grade III, IV, V, and VI used in this study, (3) the progress of elementary students gade III, IV, V, and VI ability (θ) in Mathematic based on the developed test. The data of this study were students’ scorer on final exams Elementary school 2013 collected with anchor test design. The result showed that: (1) Standard Error of Equating (SEE) within the linear method was 0.607 and ekipercentile method was 0.505116, while Root Mean Square Different (RMSD) within the meanmean was 0.394 and meansigma method was 0.344. The differences between the value of SEE and RMSD was very little, therefore the meaningfulness needed further study; (2) the Mathematic achievement test for Grade III, IV, V, and VI had location in the middle category within the values respectively -0.070, -0.228, -0.324, -0.430, the good mean of slope 0.537, 0.558, 0.639, 0.644, and items fit statistics minimum 0.077, 0.052, 0.055, 0.198; (3) the four equation method provided information that θ increased in accordance with the increased of the class rank, with mean of the θ students Grade III, IV, V, and VI as 0.104, -0.076, 0.004, dan 0.173. But the increased of the θ value was relatively small, in that its meaningfulness also required further study. Keywords: equating method, mixed-model test