KOMPARASI ESTIMASI RERATA SKOR RESPONDEN ANTARA ITEM SAMPLING, EXAMINEE SAMPLING, DAN MULTIPLE MATRIX SAMPLING

Abstract

The purpose of this study is to find out which is the most accurate sampling design in estimating the mean score of respondents between item sampling, examinee sampling, and multiple matrix sampling. The population was Jakarta students' answers on the National Examination in 2013 on Bahasa Indonesia subject. This study using three sampling designs that named: item sampling, examinee sampling, and multiple matrix sampling. To obtain 30 averages for each designs, the data is empirically replicated 30 times, and then analized statistically. Although all data for each designs have normal distribution but it was not homogeneous, so hypothesis must be tested using the Kruskal-Wallis test. The result is there are significant differences between the three kind of the sampling design. Then the test followed by Mann Whitney test. The result show that the examinee sampling is more accurate than the item sampling, multiple matrix sampling is more accurate than item sampling, and multiple matrix sampling is more accurate than the examinee sampling.