Page 125 - DEDU504_EDUCATIONAL_MEASUREMENT_AND_EVALUATION_ENGLISH

Unit 9 : Test Standardization

9.4.6 Reliability Coefficient Notes
This requires correlating scores on two equivalent forms of the same test, given simultaneously
to the same group of students using the same procedure. This measure is the coefficient of
realibility. The following types of coefficients are commonly used.
Retesting Coefficient : When only one form of test is available, it is given to the group of pupils
twice under similar testing conditions. Retesting coefficient is the correlation coefficient between
two sets of scores. However, second administration should neither follow the first too quickly to
avoid significant increase of scores that may result from memory, nor be delayed too much lest
forgetting operates to a large extent.
Chance Half Coefficient (Split Half) : The test is given to a group of pupils and their scores are
then obtained for two halves of the test. Two halves can be made as :
(i) odd and even numbered items; and

(ii) obtaining separate scores on items 1, 4, 5, 8, 9, 12, 13 etc. and on items 2, 3, 6, 7, 10, 11 etc. to
equalise the difficulty of the two halves when items are scaled in a scaled order of difficulty.
Correlation coefficient obtained between the two sets of scores indicates the degree of
conformance between the two chance halves of the test. Reliability coefficient of the test is
then found by estimating the correlation by using Spearman-Brown Prophecy formula :
r
2½½
r = 1+ ½½
12
r
Foot Rule Coefficient : This may be an underestimate but never overestimate of the reliability
coefficient. It is not the most accurate method. It requires use of three facts and measures from
the test in a simple formula - the arithmetic means, standard deviation of scores and the
number of items in the test. Owing to sufficient accuracy and simplicity, this method is
recommended for use by teachers in estimating reliability of their informal objective
examinations. The formula used is given below :
xk )
( – x
rH = k ( 2 ) SD

where rH = coefficient correlation;

x = mean;
k = number of items in the test; and

SD 2 = variance (Standard deviations squared).
Estimate of reliability coefficient often results in high or low test reliability. It must be based on
known and appropriate range of ages or grade placement of pupils if it is to mean what it
purports to mean. Hence reliability coefficient is neither an entirely adequate device, nor for that
matter the only method of indicating the internal consistency of a test.
Standard Error of Measurement (SEM) : The other popular device by which test reliability can be
estimated is the standard error of measurement. Standard error indicates the degree of accuracy
existing in the test score, obtained for each pupil on a test.
Here the accuracy refers” to magnitude of sampling errors. Since SEM isnot ecTed by range of
talent of the pupil group on which it is based (as in reliability coefficient), it is recognised as a
more concrete way of indicating test reliability.
Adequacy and Objectivity in Test Reliability : “Adequacy” is the degree to which test samples
sufficiently widely into the subject so that the resulting scores are representative of relative total

LOVELY PROFESSIONAL UNIVERSITY 119