Page 91 - DEDU504_EDUCATIONAL_MEASUREMENT_AND_EVALUATION_ENGLISH

Unit 7 : Reliability – Types , Methods and Usability

Notes
Odd Even
1 2
3 4

5 6
7 8
“ “
“ “
“ “
“ “
“ “
99 100

Because, the second question is more difficult than the first one, the fourth is more difficult
than the third one; the sixth question is more difficult than the fifth one, and so on. Therefore,
the even-numbered test will be more difficult than the odd-numbered test. This problem is
solved by using the following way :

First Part Second Part

1 2
4 3
5 6
8 7
9 10

Thus, a 10-question test has been divided into 1, 4, 5, 8, 9 and 2, 3, 6, 7, 10.
The correlation coefficient (r) between the two will be its reliability. To know the reliability
of the whole test, the Spearman-Brown prophecy formula is used, which is are as follows :

2r 1/2'1/2
r
11 =
+
1 r 1/2'1/2
Where, R = Reliability coefficient of the whole test
11
R = Reliability coefficient of one-half test
1/2’1/2
Supposing, the reliability coefficient of two parts of a test is r = .75, on this basis, the
reliability coefficient of the entire test will be :
×
2.75 1.50
r = = = .9 (approx)
+
1.75 1.75
That is, reliability coefficient of the entire test will be .9 (approx).
Limitations
(a) A limitation of this method is that chance error can arise in the test, which increases
the amount of r in two parts.
(b) It is not possible to use these tests in speed tests because the difficulty level of
questions there is equal. Nunnly too has said : “It Is particularly misleading to use
the sub-divided test method on a highly speeded test.”

LOVELY PROFESSIONAL UNIVERSITY 85