ABSTRAK

WASIS: Model Penskoran Partial Credit pada Item Multiple True-False Bidang Fisika. Disertasi. Yogyakarta: Program Pascasarjana Universitas Negeri Yogyakarta, 2009

     Penggunaan format item pilihan ganda dan penskoran dikotomus untuk mengukur kemampuan di bidang fisika menunjukkan sejumlah kelemahan. Penelitian ini merupakan salah satu usaha untuk mengatasi kelemahan tersebut. Tujuan penelitian ini menghasilkan model penskoran politomus untuk respons item multiple true-false, sehingga dapat menaksir kemampuan di bidang fisika lebih akurat.

    Pengembangan penskoran mengikuti tahapan Four-D model dan diuji akurasinya melalui penelitian empiris dan simulasi. Penelitian empiris untuk menemukan model pengategorian partial credit yang secara nyata mampu menaksir kemampuan lebih akurat dibandingkan model pen­skoran dikotomus. Penelitian empiris menggunakan 15 item multiple true-false yang diambil dari soal Seleksi Penerimaan Mahasiswa Baru (SPMB) atau Ujian Masuk Perguruan Tinggi Negeri (UMPTN) tahun 1996-2006, dan dikenakan pada 410 mahasiswa baru Fakultas Matematika dan Ilmu Pengetahuan Alam (FMIPA) Univer­sitas Negeri Surabaya angkatan tahun 2007. Respons peserta tes diskor secara partial credit dengan tiga model, yaitu: (i) memperhatikan kompleksitas setiap option (pembo­botan), yang disebut penskoran Partial Credit Model (PCM) I; (ii) tanpa memperhatikan pembo­botan, yang disebut PCM II; dan (iii) berdasarkan tingkat kesukaran tiap option, yang disebut PCM III. Respons peserta tes juga diskor secara dikotomus. Empat hasil penskoran di atas dianalisis dengan program Quest untuk menda­pat­kan taksiran tingkat kesukaran item (δ) dan taksiran kemampuan peserta (θ). Dua parameter tersebut digunakan untuk menen­tukan nilai fungsi informasi tes dan kesalahan baku taksiran. Semakin tinggi nilai fungsi informasi dan semakin kecil kesalahan baku taksiran, berarti semakin akurat model penskoran tersebut dalam menaksir kemampuan. Penelitian simulasi untuk memantapkan hasil penelitian empiris, dengan menguji pengaruh jumlah kategori, panjang tes, dan ukuran sampel terhadap akurasi taksiran kemampuan. Penelitian simulasi menggunakan data bangkitan berdasarkan parameter empiris (δ dan θ) dan parameter komputasi (δ, θ, ukuran sampel (N), dan pan­jang tes). Pembangkitan data simulasi menggunakan program statistik SAS dan akurasi taksirannya dianalisis dengan metode root mean squared error (RMSE). Semakin rendah nilai RMSE, semakin akurat taksiran kemampuan yang dihasilkan melalui simulasi.

      Hasil penelitian ini menunjukkan: (i) Penskoran model partial credit dengan pembobotan mampu menaksir kemampuan lebih akurat dibandingkan bila tanpa pem­bobotan, apalagi bila dibandingkan dengan penskoran dikotomus; (ii) Semakin banyak jumlah kategori dalam penskoran partial credit, semakin akurat taksiran kemam­pu­an yang dihasilkan;  (iii) Semakin panjang tes, semakin akurat taksiran kemampuan yang dihasilkan; (iv) Ukuran sampel tidak menunjukkan pengaruh yang signifikan terha­dap hasil taksiran kemampuan, tetapi jumlah sampel di atas 300 responden menunjukkan hasil taksiran yang lebih stabil; dan (v) Analisis dikotomus per option pada item multiple true-false dapat digunakan untuk melacak terjadinya miskonsepsi, mengidentifikasi berba­gai keterampilan, serta mengenali sumber kesalahan dalam penyelesaian per­masalahan fisika. Berdasarkan hasil penelitian di atas, direkomendasikan penggunaan item multiple true-false dan penskoran politomus model partial credit dengan pembobotan pada penilaian kemampuan fisika, dan secara umum pada penilaian kemampuan bidang lain yang memiliki respons bertingkat.

ABSTRACT

WASIS: The Partial Credit Scoring Model for the Multiple True-False Items in Physics. Dissertation. Yogyakarta: Graduate School, Yogyakarta State University, 2009

     The use of the multiple choice item format and dichotomous scoring to measure abilities in physics shows several weaknesses. This study is an attempt to overcome the weaknesses. This study aims to produce a polytomous scoring model for responses to multiple true-false items in order to result in a more accurate estimation of abilities in physics.

     The scoring development adopts the Four-D model and its accuracy is assessed through empirical and simulation studies. The empirical study was carried out to find the partial credit categorizing model that is significantly capable of estimating testees' abilities more accurately than the dichotomous scoring model. The empirical study employed 15 multiple true-false items taken from the tests for SPMB (Seleksi Penerimaan Mahasiswa Baru; New Students Entrance Selection) atau UMPTN (Ujian Masuk Perguruan Tinggi Negeri; State Univer­sity Entrance Examination) from 1996-2006 and they were administered to 410 new students enrolled in 2007 in FMIPA (Fakultas Matematika dan Ilmu Pengetahuan Alam; Faculty of Mathematics and Science) of Surabaya State University. The testees' responses were scored using the partial credit with three models, namely (i) the model taking account of the complexity of each option (weighting), called the scoring with the Partial Credit Model (PCM) I; (ii) the model not taking account of weighting, called PCM II; and (iii) the model based on the difficulty level of each option, called PCM III. The testees' responses were also dichotomously scored. The results of the four scoring models were analyzed using the Quest program to obtain the estimation of the item difficulty level (δ) and that of the testees' abilities (θ). The two parameters were used to find the information function of the test and the standard error of the estimation. The higher the information function was, the smaller the standard error of estimation would be, indicating that the scoring model was more accurate in estimating testees' abilities. The simulation study was conducted to strengthen the results of the empirical study by testing the effects of the number of the partial credit categories, test length, and sample size on the accuracy of the ability estimation. The simulation study employed the generated data based on the empirical parameters (δ and θ) and the computational parameters (δ, θ, sample size (N), and test length). The generating of the simulation data used the SAS statistical program and the estimation accuracy was analyzed by using the root mean squared error (RMSE) method. The lower the RMSE value was, the more accurate the ability estimation obtained through the simulation would be.

       The results of the study show the following: (i) The scoring with the partial credit model with weighting is capable of estimating abilities more accurately than that without weighting, let alone the dichotomous scoring; (ii) The more the number of the categories in the partial credit scoring is, the more accurate the result of the ability estimation will be;  (iii) The longer the test is, the more accurate the result of the ability estimation will be; (iv) The sample size does not show a significant effect on the result of the ability estimation, but the number of sample members above 300 respondents shows a more stable result of the estimation ability; and (v) The dichotomous analysis per option on the multiple-false items can be used to trace the possibility of misconception, identifying a variety of skills, and identifying sources of errors in solving physics problems. Based on the results of the study, it is recommended that the multiple true-false items and the polytomous scoring with the partial credit model with weighting should be used in the assessment of abilities in physics, and generally in the assessment of other abilities with stratified responses.

kirim ke teman | versi cetak