Research Setting Research Population and Sample
33
Besides, the school did not allow the research to be extended, since the students are now in the twelfth grade.
The second method is the split half method. Cronbach 1947 describes this method as splitting or dividing the measurement scale into two matched
halves. The teachers then need to calculate the correlation of these two tests results. Hyland 2003 mentions that a test reliability is higher when the result is
closer to 1. Seeing the nature of spilt half method, it is best applied to multiple choices. Therefore, the researcher did not use this method.
The last method is inter-rater reliability, in which the students’ writings are assessed by several raters Weigle, 2002. Similar to split half method, the
teachers then calculate the correlation of the test score given by the raters Lado, 1961. Hudges 1989 also adds that even though the writing test tends to be
subjective, there will be fewer problem when the raters have agreement on the ratings and the way they assess.
The researcher applies this method to ensure the test reliabilty in the research. The raters are the researcher and one student of English Language
Education Program 2010. She was chosen as the rater because a rater should have at least the basic knowledge of assessing writing and also get explanation of the
rubric used Wang, 2009. After the raters assessed the students’ writing, the researcher calculates the score by using SPSS ver.20 Full Version, applying
Pearson statistical test. This test aims to measure the relation between two variables the score from rater 1 and rater 2 Norusis, 2000. Presented below is
34
Table 3.1 presents the of reliability coefficient of the test used in the research. For the complete data of the score given by the raters, please refer to APPENDIX 5.
Table 3.1.Table of Correlation Coefficient Calculation Result
Group Correlation Coefficient
Pre-test Post-test
Experimental Group 0.77
0.84 Control Group
0.84 0.88
The table above shows that the reliability coefficient of the pre-test for experimental group is 0.77 and for the control group is 0.84. While for the post-
test, the reliability coefficient for experimental group is 0.84 and 0.88 for the control group. Lado 1961 explains that higher coefficient closer to 1.00
indicates higher reliability. Taylor 1990 adds that 0.36 until 0.67 is the modest or moderate correlation. Therefore, it can be concluded that the result of
correlation coefficient is strong, as it exceeds the moderate limit. Other than reliability, there is another thing that is also an essential requirement of an
effective test Heaton, 1975.