invalid items, the researcher did the second step of pre-test try out calculation. In this step 3 items were found invalid. They were 6, 13 and 39 the items number. Then, the
researcher came to the third step and found no single item which was considered invalid. Finally the total valid items in pretest try out were 20 items. See appendices
6,7 and 8 While in the first step of the post-test try out, there were 17 items considered invalid.
They were the items number 4, 5, 7, 12, 13, 14, 17, 18, 22, 27, 29, 30, 31, 33, 34, 38, dan 40. After dropping those invalid items, the researcher did the second step of post-
test try out calculation. In this step 3 items were found invalid. They were the items number 2, 8, dan 19. Then, the researcher came to the third step and found no single
item which was considered invalid. Finally the total valid items in post-test try out were 20 items. See Appendices 9, 10 and 11.
2. Reliability of Test
Reliability refers to the consistency of test scores. Besides having high validity, a good test should have high reliability too.
Alpha formula will be used to know reliability of test K- R.20.
13
R11 = -
R11 = The reliability coefficient of items
k = The number of item in the test
p = The proportion of the students who give answer the item 1
q = 1-p
= Sum of p time q S
2
= Variance of the total score
13
Sugiono, Op Cit. p. 132.
The criteria of reliability test are : 0.80-1.00
= Very high reliability 0.60-0.79
= High reliability 0.40-0.59
= Medium reliability 0.20-0.39
= Low reliability 0.00-0.19
= Very low reliability.
14
From the criteria of reliability above, it can be drawn a conclusion that the result of reliability for pre-test has a high reliability since it amounts to 0.77 and the result of
reliability for post-test has a high reliability because it amounts to 0.77. It means that reliability of the test in the research are reliable. See Appendices 12 and 13
I. The Data Analysis 1. Normality Test
To analyze the data, the researcher needs to test the data distribution, whether it is normal or not. The normally test is used to measure whether the data in the
experimental class and control class are normally distributed or not. In this case, the researcher will use Lilliefors test as follows :
The hypothesizes for the normality test were formulated as follows: H
o
: The data are normaly distributed. H
a :
The data are not normally distributed.
14
Sugiono, Ibid. p. 184.