Measurement, Reliability &validity

An Introduction to the Scale of
Measurement
Measurement is assigning numbers to
observations in such a way that the numbers are
amenable for analysis.
The number represents the property being
studied.
All the variables can not be measured in one
scale.
There are four scale of measurement viz;
Nominal, Ordinal, Interval and Ratio scale of
measurement.
1.Nominal Scale
The operation is partitioning the objects, persons
or characteristics in to mutually exclusive sub
classes and the relation between the members
of the class is equivalence (=).
Classifying a group of persons into male and
female and assigning numbers as 0 and 1.
Automobile license plate numbers are other
examples.
Numbers only denote do not connote.
2.Ordinal Scale
Objects of one category are not different( as
measured in nominal scale) but they stand in
some kind of relation amongst them. < or > ,=
We rank objects/ statements by giving counting
numbers.
These numbers are not isomorphic to the
system of arithmetic. The successive difference
are not same.
Expl 1 : Ranking of individuals with respect to
levels of income, Exam results: passed in
1st,2nd or 3rd class.etc
Likert Scale numbers are often considered as

measured in ordinal scale
Listed below are some of the services that

may influence the choice of a bank.
Please rate the following in a 5 point scale
1- Least Important
5- Most important
We normally rank the attributes/variables
in a Likert Scale.
Ranks are not isomorphic to the system of
arithmetic.
Example 2 (Contd)
Services
1.
Courteous
service
2.Compete
nce of staff
3.Fast
action on
complaints
5.No of
branches
6.Availabilit
y of ATM
services
7.Financial
strength of
the bank
3.Interval Scale
An interval scale is characterized by a
common and constant unit of
measurement, but with an arbitrary zero.
Consider the measurement of temperature
in two scales e.g. F & C
The two scale conform to the linear
transformation such as
F = 9/5 C + 32
Two points to observe

1. Zero points in the two scale are arbitrary.
2.The ratio of difference between reading on one
scale is equal to that of the other scale. But the
ratio of scale value is not equal due to the
arbitrary zero.
E.g. Cen 0 10 30 100
Fer 32 50 86 212
30-10/10-0 = 86-50/50-32 = 2
But 0 : 32 # 10:50 # 30: 86 # 100 : 212
4.Ratio Scale
A Ratio Scale of Measurement has all the
characteristics of interval scale with true
zero point at its origin.
This scale is isomorphic to the system of
arithmetic (with a true zero).
It has known ratio of any two interval and
known ratio of any scale value
Consider the following.

Measurement of area is done with acre or
hectare which are in ratio scale.
Area in Ha
2
4
6
8
10
Area in acre
5 10
15
20
25
Known ratio of difference
6-4/ 4-2 = 15 - 10/ 10 - 5 = 1
Known ratio of scale value
2:5 = 4:10 = 6:15
Scale of Measurement and Statistical

analysis
The numbers assigned by nominal and ordinal
scales known as non-metric & are not amenable
for some statistical analysis.
Interval and ratio scales are known as metric
scale of measurement and the variables
measured in this scale are amenable for
statistical tools.
However, variables measured in all the scale of
measurement are used in social science
depending on the objectives & mode of use.
Reliability and Validity

After assigning numbers to the
objects/events, properties according to
rules the researcher asks two questions.
1. What is the reliability of the measuring
instrument?
2.What is its validity?
These two are important since the

numbers represent the phenomena under
study.
Reliability
Reliability in the context of measurement is
based on the probability of errors.
It implies: If the same object or property is
measured again and again with the same
measuring instrument we get similar result.
To the extent that the errors are present in a
measuring instrument, to that extent the
instrument is unreliable.
Thus, reliability is the relative absence of error of
measurement in a measuring instrument.
Reliability Measures
A Measure of reliability is the proportion of the
"true" variance to the total obtained variance of
the data yielded by a measuring instrument.
Alternatively it is the proportion of the error
variance to the total obtained variance yielded
by a measuring instrument subtracted from the
index of "One".
E.g. 1. R 2 = ESS/TSS or 1- RSS/TSS
Cronbach's Alpha - Measures dimension.
Validity
Validity concerns an enquiry in to the reality of a
variable and theoretical consistency.
Consider the following two sets of varables
1. Gender, Domicile, Length, Weight, income etc.
2.Personality,Brand preference, loyalty, awareness
For the former there are specific measures, but for
the later there are indirect ways to measure.
A measurement may be reliable but may not
have validity.
Types of Vilidity
Researchers refer to three types of validity
Viz Content, Criterion & Construct validity
a ) Content validity
It is the extent to which a measuring
instrument provides adequate coverage of
the concept/topic/ entities.
E.g . A proxy variable representing a qualitative
variable
A representative sample in a sample study,
Criterion validity
It relates to efficiency of a model in
estimation and prediction.
These could be achieved with the
use of different specification with
respect to variables or the use of
different statistical models or
relationships.
The model having least error is
supposed to have criterion validity.
Construct Validity
It is more complex, subjective & abstract and
unites psychometric notions with theoretical
notions.
It is not only validating the relationship, but
one must try to validate the theory behind it
e.g. in a dependency relationship Y and a few
Xi s are related and the researcher asks why
such a relationship may exit?
A Priori Reasoning has to be given which
concerns theory, logical judgment &
researchers articulation.
A priori reasoning
Two variable cases
(i)
Demand of X
(ii)
= f (Price of X )
Income Tax Rev. of

GoI
(iii) Agricultural
Production
= (Industrialization)
(iv) Production of a
group of small scale
industries
= f (Working Capital)
(v)
= f (Area under cultivation)
Industrial Production = f (Employment in Organised Sector)
A priori reasoning.
# More than two variables
(i)
Demand of X
= f (Price of X, Price of Substitute, Adv.

Expenditure)
Income Tax Rev. of

GoI
(iii) Agricultural
Production
= (PCI, Literacy Rate, Industrialization)
(iv) Production of a
group of small scale
industries
= f (Working Capital, Proximity to

market, Ratio of Number of Technical
Persons/Total Employees)
(v)
= f (Expenditure in R & D, Employment

in Organised Sector)
(ii)
Index of Industrial
Production
= f (Area under cultivation, % of area

irrigated, fertilizer use)
Construct validity is addressed with

respect to the following
Specification of proxy variables or quantification of
qualitative variables.
Whether the researcher is near the property being
measured while specifying proxy variables.
Specification of relationship between dependent and
independent variables (proper articulation with valid
logic)
Redundant relationship with no a priori reasoning should
be avoided
Specification of functional forms e.g. Linear,
Quadratic, multiplicative, logarithmic etc.

Measurement, Reliability &validity

Uploaded by

Copyright:

Available Formats

Measurement, Reliability &validity

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Measurement, Reliability &validity

Uploaded by

Copyright:

Available Formats

An Introduction to the Scale of

Likert Scale numbers are often considered as

Listed below are some of the services that

Two points to observe

Consider the following.

Scale of Measurement and Statistical

Reliability and Validity

These two are important since the

Income Tax Rev. of

= f (Area under cultivation)

Industrial Production = f (Employment in Organised Sector)

= f (Price of X, Price of Substitute, Adv.

Income Tax Rev. of

= (PCI, Literacy Rate, Industrialization)

= f (Working Capital, Proximity to

= f (Expenditure in R & D, Employment

= f (Area under cultivation, % of area

Construct validity is addressed with

You might also like