Variables, Measurement, and Scaling Technique - Research Methodology - page 42

pairwise comparison is taken up. As a result, if
there are

5.3.2 Errors in Scaling Techniques

numbers of objects to be compared,
then one needs to compare N C 2 pairs. If

N

is very
large, then the number of combinations will also
be very large. On the other hand, in method of
rank order, a respondent orders or ranks all
objects according to the specified criteria into

N If a respondent is not very careful while rating or

ranking, errors may occur. Mainly there could be
three types of error: error of leniency, error of
central tendency

.The
error of leniency takes place when respondents
are not so serious in taking the job of rating/
ranking on their own hands. On the other hand,
the error of central tendency occurs when
respondents are very much reluctant to provide
the highest/lowest rate/rank. Thus, in grading
students, a teacher may be reluctant in judging
the answer script thoroughly but put the grade
leniently. Contrarily, if a teacher never wishes to
provide the highest rank or the lowest rank but
rather puts the average rank, then errors of central
tendency occur. Errors of systematic bias or halo
effects occur when a respondent carries or has
preconceived ideas about the objects under rating.
In most of the cases, the raters carry a generalized
impression from one subject to another.

,and error of halo effect

a ,

b . In many situations,

respondents are asked to rank a particular object
or to rank the first three and so on. Thus, the need
of ranking all the objects under consideration
doesn't arise at all. The problem with this type
of method is that the carelessness or too much
sensitivity on the part of the respondent may lead
to bad information or time-consuming affairs,
respectively. When there are a large number of
objects for which the selections are to be made, it
becomes very difficult to use the previous two
methods. Instead, one can use the method of
successive intervals. The large number of objects
is sorted to different groups formed by taking
interval scales.

, c ... or 1, 2, 3

...