Page 221 - DEDU504_EDUCATIONAL_MEASUREMENT_AND_EVALUATION_ENGLISH

Unit 18 : Marking System : Need, Problems, Components

(iv) Informal derived scores : Relative ranks and letter marks on a test (A, B, C, D, E, F) are Notes
other types of derived scores.

18.5 Components of Marking System

Numerical basis for assigning marks should include different aspects like home task, project, test
scores and class-room contribution. Weightage to each component may be worked out on the
basis of Mean and S.D. of component score, for getting the combined score. However, precise
weighthing of components on numerical basis is not crucial to the quality of scores assigned.

18.6 Techniques of Marking

Scaling and Equating
Scaling is a technique that standardises raw scores or marks from one scale to another. Two raw
scores may be described equivalent and assigned the same degree of excellence in relation to
some relevant group. Equivalence means that when raw scores are normally distributed, the two
definitions produce identical results (Harper). When two distributions of raw scores are not
normal, the calibrated scores will be different for these two approaches. Two raw scores are
defined as equivalent and all therefore are translated into the same scaled score if they are at the
same distance from the means of their distribution, in terms ot standard deviations of their
x – m
distribution, in terms of standard deviations of their distribution, i.e. (x = score, m = mean,
S.D.
and S.D. = standard deviation). Two raw scores are considered equivalent and are therefore
translated into the same scale score if they are both exceeded oy the same proportion of examinees.
For example, if 10% of the students receive a mark of 55 or higher on an examination P, and a
mark of 63 or higher on examination Q, these two raw marks (55 and 63) should be awarded the
same scaled marks or grade. Calibration of any marks will control enough of the factors to make
the scaled marks or grades much more valid than the raw marks. Since in many cases we deal
with population not a sample (any students who sat for examination), scores need not necessarily
be distributed according to normal curve. The following methods can take care of non-normally
distributed groups.
Use of Linear or Normalised Method
Linear-scale transformation produces a set of scores whose shape distribution is identical with
that of the raw scores, whereas normalised scale transformations force a non-normal distribution
of raw scores into a normal distribution of scaled scores.
Which of the two should be used ? Draw a graph of the raw scores awarded by several examiners
and smoothen the curves. If distribution is approximately normal, any of the two methods as far
convenience can be used. But if they are skewed or very irregular, find out why they are not
normal. If it is because of sampling, use the linear scale method. If it is due to examiner who
skewed the marks or due to the peculiarity of the test that has produced non-normal distribution,
use the normalised scaling. Situation second is more common than the first one. Advantage of
normalising is that it makes the scaled scores strictly comparable at all levels, though linearly-
scaled scores may not have equivalent distribution, the linearly scaled marks supposedly cannot
legitimately be added or compared directly.

Equating
Sometimes when we suspect that a particular group is not really a representative group, we may
prefer to have a population (not sample), e.g. “all students of this years Biology class that I have
taught are to be judged on the same standard as those of last year”. This is called equating
because the final grades or marks are based on this year’s statistics and on past statistics also.

LOVELY PROFESSIONAL UNIVERSITY 215