Claim Missing Document
Check
Articles

Found 4 Documents
Search
Journal : JEC (Jurnal Edukasi Cendekia)

EFEK TEST LENGTH DAN NUMBER OF EXAMINEES TERHADAP TEST RELIABILITY Samritin Samritin
JEC (Jurnal Edukasi Cendekia) Vol 1 No 1 (2017): Februari
Publisher : Lembaga Penelitian dan Pengabdian Masyarakat FKIP UMB

Show Abstract | Download Original | Original Source | Check in Google Scholar | Full PDF (812.348 KB)

Abstract

Reliabilitas merupakan salah satu properti tes yang berkaitan dengan kesalahan atau konsistensi pengukuran. Kesalahan pengukuran dalam teori tes klasik dihasilkan perbedaan antara skor tampak dengan suatu skor yang sesungguhnya-tidak tampak (skor murni). Reliabilitas suatu tes terlihat dari koefisien reliabilitasnya. Koefisien reliabilitas dapat diestimasi dengan beberapa cara, salah satunya tergantung pada jenis tes. Hasil estimasi koefisien reliabilitas dapat dipengaruhi oleh beberapa faktor, antara lain panjang tes dan heterogenitas kemampuan peserta tes. Hasil penelitian simulasi ini menunjukkan bahwa semakin panjang suatu tes, koefisien reliabilitasnya semakin tinggi. Sebaliknya semakin pendek suatu tes, koefisien reliabilitasnya semakin rendah. Ditemukan juga bahwa tes yang memiliki tingkat kesukaran tinggi, koefisien reliabilitasnya lebih tinggi dibandingkan dengan koefisien reliabilitas tes yang lebih pendek. Dalam penelitian ini tidak ditemukan pola perubahan koefisien reliabilitas akibat peningkatan atau pengurangan banyak peserta tes.
Pengaruh Sample Size terhadap Kestabilan Item Parameter Estimate pada Item Response Theory Model Tiga Parameter Logistik Samritin Samritin
JEC (Jurnal Edukasi Cendekia) Vol 1 No 2 (2017): Agustus
Publisher : Lembaga Penelitian dan Pengabdian Masyarakat FKIP UMB

Show Abstract | Download Original | Original Source | Check in Google Scholar | Full PDF (666.44 KB)

Abstract

This study aims to determine the stability of the parameter estimates of the test items on the modern test theory (IRT) for the three-parameter logistics model (3PL). This study uses a simulation method. Simulation data was generated using the WinGen for Windows program. Estimation stability analysis was carried out using the correlation between true parameters with estimate parameters and RMSE. The results showed that for sample sizes 500, 750, and 1000, the correlation coefficient (ρ) was 0.901, 0.988, and 0.992 respectively with RMSE 0.901, 0.351, and 0.273 for the difficulty parameter (b), correlation coefficient (ρ) of 0.746, 0.807, and 0.830 with an RMSE of 0.746, 0.304, and 0.281 for the power of difference parameter (a), the correlation coefficient (ρ) of 0.642, 0.667, and 0.679 with an RMSE of 0.642, 0.057, and 0.056 for the pseudo-guessing parameter. Based on the results of data analysis it was concluded that the larger the sample size, the more stable the item parameter estimates
EFEK NUMBER OF ITEM DAN SAMPLE SIZE TERHADAP AKURASI ABILITY ESTIMATION PADA MODEL IRT TIGA PARAMATER LOGISTIK Samritin Samritin
JEC (Jurnal Edukasi Cendekia) Vol 2 No 1 (2018): Februari
Publisher : Lembaga Penelitian dan Pengabdian Masyarakat FKIP UMB

Show Abstract | Download Original | Original Source | Check in Google Scholar | Full PDF (1001.245 KB)

Abstract

This study aims to determine the effect of the number of items and sample size on the estimated accuracy of the 3PL IRT Model. This research is a simulation research that uses three variations of the number of items, namely 30, 40, and 60 with three variations of the sample size, namely 500, 750, and 1000. The research data is obtained from the results generated by a computer application program. The data obtained were analyzed descriptively using the correlation coefficient, mean square difference (MSD), and bias. The results of this study indicate that based on the correlation coefficient the highest accuracy was obtained at N=1000 with n=60 but the lowest MSD was obtained at N=750 and n=60, the lowest bias was at N=500 and n=60. Based on the results of the MSD analysis and the bias of this study it was concluded that sample size has an unadjusted effect on estimation ability. This study also found that the more items tested, the more accurate the estimation of the ability of the test takers
KALIBRASI TES CAMPURAN DIKOTOMUS 2PLM DAN POLITOMUS GRM MENGGUNAKAN PROSEDUR GRM DAN GPCM Samritin Samritin
JEC (Jurnal Edukasi Cendekia) Vol 2 No 2 (2018): Agustus
Publisher : Lembaga Penelitian dan Pengabdian Masyarakat FKIP UMB

Show Abstract | Download Original | Original Source | Check in Google Scholar | Full PDF (635.714 KB)

Abstract

Item parameter estimation is an important step in test development. Estimation of test parameters must pay attention to the suitability of the response characteristics with the calibration procedure to be used. This study has simulated mixed test response data of 35 dichotomous items of 2PLM with 5 polytomous items of GRM. Response data was generated using a sample size of 1500 examinees. Response data were analyzed descriptively from the aspect of correlation coefficient, root mean square difference (RMSD), mean absolute difference (MAD), variance, and bias. The results of this study indicate that the estimation of the slope and threshold parameters of the dichotomous test model 2 logistic parameters using GRM calibration produces correlation coefficients, RMSD, MAD, variance, and estimation bias that are almost the same as those of the GPCM calibration; parameter estimation of the slope and threshold of the GRM model polytomous test using GRM calibration produces a smaller correlation coefficient, RMSD, MAD, variance, and estimation bias compared to the GPCM calibration results; and estimation of the slop parameters and threshold of the mixed dichotomous 2 parameter logistic model (2PLM) test and the GRM polytomous test calibrated using the GRM procedure produced a higher correlation coefficient but had lower RMSD, MAD, variance, and bias compared to the results of calibration using the GPCM procedure. Based on the results of data analysis, this study concluded that the parameter estimation of the 2PLM dichotomous and polytomous GRM mixed test is more precise when calibrated using the GRM procedure.