Subject Psychology

Paper No and Title Paper 4: Applied Psychometrics

Module No and Title Module 2: Measurement & Scaling

Module Tag PSY_P4_M2

TABLE OF CONTENTS
1. Learning Outcomes
2. Introduction
3. Scales of Measurement: An Introduction
4. Nunnally’s Fundamentalist view of Measurement Scales
5. Types of Measurement Scales: Continuous and Discrete
6. Types of Measurement Scales
6.1 Nominal Scale
6.2 Ordinal Scale
6.3 Interval Scale
6.4 Ratio Scale
7. Tabular Representation of Measurement Scale Types with Major Attributes
8. Measurement Scales in Psychology
9. Summary

PSYCHOLOGY PAPER 4: Applied Psychometrics

MODULE 1: Measurement and Scaling
PSYCHOLOGY PAPER 4: Applied Psychometrics
MODULE 1: Measurement and Scaling
1. Learning Outcomes
After studying this module, you shall be able to:

 Understand types of scales used in psychological assessment.

 Understand Nunnally’s “Fundamentalist” View of Measurement Scales.
 Differentiate between Continuous and Discrete Scales
 Define and Differentiate among Nominal, Ordinal, Interval and Ratio scale.

2. Introduction
People differ in their physical and psychological attributes. Though it is easy to measure physical
characteristics of people like height, weight due to the availability of standardized scales or
standard units; measuring psychological characteristics raises a fundamental problem as no
similar external standard exists. There is no perfect ability, personality or attitude measure with
which a standard response can be obtained. A measure of intelligence does not have the same
mathematical status as does a meter scale.

During recent years in psychology there has been much talk about the different possible types of
measurement scales, and there has been much soul-searching about the types of scales
characterized by different types of psychological measures. This module deals with ‘Scales of
Measurement’ which is the basic underlying component of psychological assessment. Scales
enable us to transform the responses of people into quantitative scores which form the basis of
measuring attributes/characteristics. A simplified, conventional and most widely accepted
classification of measurement scale comprises of four types- Nominal Scale, Ordinal Scale,
Interval Scale and Ratio Scale.

Nominal Scale is the most rudimentary type of measurement scale in which numbers are coded
into categorical information. The next level of measurement is Ordinal Scale. In this type of
measurement, a set of objects or people is ordered from “most” to “least” with respect to an
attribute but there is no indication of how far apart the objectives are with respect to the attribute.
An Interval Scale is one in which the numbers are rank-ordered and it is known how far apart the
objects are from one another with respect to the attribute. The Ratio scale provides the maximum
information about the numbers because in addition to having all the properties of nominal, ordinal
and interval scale, it also has a true zero point.

3. Scales of Measurement: An Introduction

According to N.R. Campbell, Measurement, in the broadest sense, is defined as the assignment of
numerals to objects or events according to rules. The fact that numerals can be assigned under
different rules leads to different kinds of scales and different kinds of measurement. Stevens, in
his earlier writings, put forward the view that assignment of numbers or numerals under different
rules or conventions leads us to different kinds of scales of measurement.

PSYCHOLOGY PAPER 4: Applied Psychometrics

MODULE 1: Measurement and Scaling
According to Stevens (1951):

“Scales are possible in the first place only because there exists an isomorphism between
properties of the numeral series and the empirical operations that we can perform with the
certain aspects of objects. This isomorphism, of course, is only partial. Not all the properties of
number and not all the properties of objects can be paired off in a systematic correspondence.
But some properties of objects can be related by semantic rules to some properties of the numeral
series. In dealing with different aspects of objects, we can invoke empirical operations for
determining equality (the basis for classifying things), for rank-ordering, and for determining
when differences and ratio between the aspects of objects are equal”.

Measurement scales are those tools without which a scientist is almost out of business. In the
process of measurement, data are represented in number form. Depending on the type of study
and data collected, those numbers carry different meanings. These meanings depend on the scale
of measurement used to analyze and interpret the data. It is essential to know which scales of
measurement is used in testing or research as it will lead the researcher to the decision of kind of
statistical manipulation required to analyze the data.

4. Nunnally’s “Fundamentalist’ View of Measurement Scales

Nunnally (1978, p.24) proposed “Fundamentalist” view that holds the following three
assumptions about the measurement scales:

1. There are distinct types of measurement scales into which all possible measures of
attributes can be classified.
2. Each measure has some ‘real’ characteristics that permit its proper classification.
3. Once a measure is classified, the classification specifies the types of mathematical
analyses that can be employed with the measure.

However, Nunnally and others have acknowledged that alternatives to the fundamentalist view
may also be viable for measurement scales. It has been argued by various researchers that this is a
narrow point of view that needs to be modified before it can meet the actualities of scientific
inquiry.

5. Types of Measurement Scales: Continuous & Discrete

A scale is a set of numbers (or other symbols) whose properties model empirical properties of the
objects to which the numbers are assigned (Cohen, 2001). Various types of scales exist. One way
of classifying scale is to label it on the basis of type of variable considered for the study. A scale
used to measure a continuous variable is known as “continuous scale” whereas a scale used to
measure a discrete variable is called as a “discrete scale”.

Consider an example in which a researcher needs to distinguish his subjects into categories of
gender-that is, either male or female. The categorization scale would be said to be discrete in
nature because it would not be meaningful to categorize a subject as anything, other than a male
or a female (Note: Individuals born with a gender-related genetic abnormality were excluded by
the researcher).

PSYCHOLOGY PAPER 4: Applied Psychometrics

MODULE 1: Measurement and Scaling
By contrast, a continuous scale exists when it is possible theoretically to divide any of the values
of the scale. Thus for a continuous scale, a measure of 1 or 2 is valid, and any value between 1
and 2 like 1.5 or 1.25 are equally valid. Any division on any unit on the scale produces a valid
possible measure. Examples include things like height or weight. You could have an object that
weighed 1 pound or 1.5 pounds or 1.25 pounds. It is important to note that measurement using
continuous scale always involves some error. For example, the weight of a person you measured
to be 1.5 pounds could, in reality, be 1.52 pounds. Most scales used in psychological testing are
continuous in nature and therefore can be expected to contain error. Errors will arise from the
mere use of a continuous scale; the number or score used to characterize the trait being measured
on a continuous scale should be thought of as an approximation of the “real” number. Thus, for
example, a score of 28 on some test of depression should not be thought as a precise measure of
depression but rather as an approximation of the real depression score had the measuring
instrument been calibrated to yield such a score. In such a case, perhaps the score of 28 is an
approximation of a real score of 27.8 or 29.3. In contrast to numbers or scores used to
characterize traits in continuous scales, the numbers or scores used in discrete scales are
presumed to be exact.

6. Types of Measurement Scales: Nominal, Ordinal, Interval & Ratio

A scale reminds us of a measuring instrument – a ruler, thermometer, weighing machine, and so
on -or a scale may be looked upon as a set of items, as in the Likert scale. It can also be a set of
questions so constructed that entities being measured can be systematically assigned scored or
values on the chosen scale. Assignment of such values (or scores) is related to the amount of the
attribute (or quality) and the entity process.

Measurement has been categorized with respect to the amount of quantitative information the
assigned numbers possesses. It is generally agreed that there are four different levels or scales of
measurement. Numbers at different levels or scales of measurement convey different kinds of
information. In psychometrics and in research in general, it is important to know which scales of
measurement are being employed, for the kind of scale will be one factor for determining which
statistical manipulations of the data would be appropriate. These four basic types of scales are:

 Nominal Scale (with which equality and sameness among entities are determined)

 Ordinal Scale (determines the degree or rank to which a particular attribute is possessed
by entities)

 Interval Scale (determines relative rank of the entities with respect to the degree to which
a particular attribute or characteristic is possessed by them and also how large the
difference between the entities is at the zero value of the variable attributes in absolute
sense)

 Ratio Scale ( determines all four relations, namely, equality, rank order, equality of
interval and the equality of ratio among the entities or phenomena)

These scales are actually four hierarchies of measurement procedures, the lowest in the hierarchy
being the Nominal Scale measurement and highest Ratio Scale measurement. That is why the
expression ‘levels of measurement’ has been used by some scholars for scales of measurement.

PSYCHOLOGY PAPER 4: Applied Psychometrics

MODULE 1: Measurement and Scaling
It may be pertinent to record here that the measurement of many physical quantities has
progressed from scale to scale. Initially, men knew temperature only by sensation; when things
were only a shade warmer or colder than other things, temperature belonged to the ordinal class
of scales. It became an Interval Scale with the development of thermometry and, thereafter, when
thermodynamics used the expansion ratio of gases to extrapolate the absolute zero, it became a
Ratio Scale.

The major characteristics of these classes of scales and the empirical operations that are possible
with each scale are summarized in Table 1 below followed by detail discussion of each individual
scale.

Table 1: Scales of Measurement and their Properties

Propert

Type of Scale Magnitude Equal Intervals Absolute Zero

Nominal No No No

Ordinal Yes No No

Ratio Yes Yes No

Interval Yes Yes Yes

Source: Kaplan and Saccuzzo, 2005

6.1 Nominal Scale: Nominal Scale is the most basic scale of measurement. It consists of
formulating purposeful and homogenous (that is, having logical relationship) set of classes or
categories of entities, facts or data on the basis of some trait(s), and assigning them some symbols
or numerals as a way of differentiating two or more classes of entities or data, and keeping track
of them. It may be pointed out that each category and its members are assigned the same symbols
or numerals. For example, we may assign individuals to such categories as males and females
depending upon their gender, or as social science student or science student based on the subject
they are studying, as adjusted and maladjusted individual, and so on.

Nominal scale is the least sophisticated and is considered the lowest level of measurement. At the
same time, it is basic to all the higher levels of measurement. In a nominal scale, the numbers are
simply ways to code categorical information. As a rule, a measurement is said to be nominal in
nature if a set of entities can be assigned to a set of classes that are:
 Exhaustive, that is, they include all the entities or the data in the set.
 Mutually exclusive, that is, in any case no entity or data of the set belongs to more than
one class.
 Each class is represented by a distinct symbol or label.

The only arithmetical operation applicable to Nominal scales is counting, a mere enumeration of

PSYCHOLOGY PAPER 4: Applied Psychometrics

MODULE 1: Measurement and Scaling
members in a category or a set. Further, mathematically, the basic characteristic of the nominal
level measurement is that the entities or data placed in a given category are equal to each other
with respect to the given characteristic. The relation of equality or entity between things
expressed by the sign’=’ (is equal to) has the logical properties of being reflective, symmetrical
and transitive. These three logical properties are operative among objects or symbols in the
Nominal Scale.
 Reflexivity means that every entity in a category is equal to itself. For example, c = a, for
all values of a in a given category.
 In a Symmetrical relation, if a stands in a certain position with respect to b, then b stands
in the same position with respect to a (that is, if a = b, then b = a).
 In a transitive relation, if a stands in a certain position with respect to b, and b stands in a
similar position with respect to c, then a also stands in that position with respect to c, that
is, if a = b and b = c, then a = c.

The following exercise will help to understand the Nominal Scale better:

Exercise:

In a survey done on 1,000 persons for the November 2008 Presidential Elections in Unites States
of America (USA), 500 persons voted for the Democratic Party (Barack Obama) and 400 persons
for the Republican party (John McCain), and the rest were undecided / neither. Express the
survey on the Nominal Scale.

Solution:

The voters fall in three mutually exclusive categories, that is, those who voted for the Democratic
Party (50 percent), those who voted for the Republican Party (40 percent) and those who were
undecided or voted for neither. So the sample is divided into three mutually exclusive and
exhaustive categories in which each class is represented by a distinct symbol or label. So, the
categories of Nominal Scale in this exercise are (a) Democratic Voters, (b) Republican Voters
and (c) Undecided/Neither.

6.2 Ordinal Scale: In the Ordinal Scale, the entities or the data are ranked with respect to the
degree to which a particular attribute is possessed by them. In other words, through Ordinal Scale
of measurement, we seek to determine the rank order or inequality of elements to which numbers
are assigned.

A typical example is A is greater than or better than or more useful than B, B is greater than or
better than or more useful than C, and so on. Such relations are designated by the symbol ‘>’
which means ‘greater than’ in reference to particular attributes. The ‘>’ relation (or relation of
higher-level order) has the logical properties of being irreflexive, asymmetrical and transitive.
 Irreflexivity is a logical property wherein it is not true that for any A, A > A.
 Asyymetrical means that if A > B then B > A.
 Transitivity means that if A >B and B > C, then A > C.
In other words, if the result of a marathon race is measured on the ordinal level scale, one can
infer that if a person A is faster than a person B, if person B is faster than person C, then person A
is faster than person C. This relation is maintained with regards to all the individuals in the group.

Ordinal Scales allow slightly more refined measurement than do the Nominal Scales because they
PSYCHOLOGY PAPER 4: Applied Psychometrics
MODULE 1: Measurement and Scaling
do allow for each entity being measured to determine whether that entity has more (that is ‘>’) of
the attribute in question than any other entity of the same (that is ‘=’) amount or less (that is ‘<’).
For instance, it is easier to identify and level of willing workers in contrast to unwilling workers,
rather than making comparison among workers in terms of the degree to which the trait is
possessed. An Ordinal Scale, however, does not reflect or tell how much in the absolute sense the
attribute is possessed by entities or how far apart are the entities with respect to the attribute.

A small exercise will help to understand the Ordinal Scale better.

Exercise:

Prepare an Ordinal Scale based on the polling survey exercise given above.

Solution

Since the political parties have got varying amount of votes that can be arranged in an ascending
or descending order, the Ordinal Scale is possible and will be as follows:

Choice of American Voters:

Democratic Party > Republican Party > Undecided

(50 %) (40 %) (10 %)

6.3 Interval Scale:

Interval Scale is not just characterized by ordered and ranked entities with respect to some
measured attribute, but, the distance or difference between neighboring rank or states is also well
defined and reflected. The distance between each successive interval or rank is constant in the
Interval Scale.

Interval Scale represents a level of measurement in which average of a set of measurements can
be obtained to give some meaningful answers. The examples of Interval Scale are Fahrenheit
scale of physical science and IQ scales in psychology.

On the Interval Scale, heat at 40 degree Centigrade or 104 degree Fahrenheit is not twice as hot as
20 degree Centigrade because temperature equivalent to 20 degree Centigrade on the Fahrenheit
Scale is 68 degree Fahrenheit, and 104 degree Fahrenheit is not equal to 2 x 68 degree Fahrenheit.
For further clarification, see the calculation below:

PSYCHOLOGY PAPER 4: Applied Psychometrics

MODULE 1: Measurement and Scaling
Also, when we reflect on the fact that 50 degree Fahrenheit is equal to 10 degrees Centigrade and
100 degrees Fahrenheit is equal to 38 degrees Centigrade, we can very clearly see that 38 degrees
Centigrade is not twice as hot as 10 degrees Centigrade. In the same manner, according to
Minium et al. (2001), 100 degree Centigrade is not twice as hot as 50 degree Centigrade. In fact,
according to them, 100 degree Centigrade is only 1.15 times the 50 degree Centigrade. This
problem mainly arises due to the fact that Interval Scales lack an absolute zero point, a point
where the attribute or property being measured is absent or does not exist.

For Interval-level data, there is no fixed zero point but each case has an absolute value associated
with it. This can be understood by studying the measurement of a psychological attribute using
interval scale. Suppose a researcher wishes to measure subjects’ “need for power’. With an
interval scale, the scores can be interpreted to mean that someone with a score of 25 is as
different from someone with a score of 30 as is someone with a score of 30 compared to someone
with a score of 35. The scale provides only interval-level information but the scores cannot be
compared with the reference point zero because no such state exists in which there is a complete
lack of “need for power” in an individual. So we cannot claim that a score of 35 achieved by a
subject on “need for power scale” means half the need for power when compared to an individual
having 70 point score. In order to make such a claim, there would be a need of absolute zero point
on the scale.

Though very few Interval Scales have been developed in the areas of social sciences, we can
achieve interval measurement of attributes such as (a) interest, attitude, personality, motivation
and so on, (b) status, reading interests, and so on and (c) recall, relevance or usefulness of
information for this purpose.

PSYCHOLOGY PAPER 4: Applied Psychometrics

MODULE 1: Measurement and Scaling
6.4 Ratio Scale: The ratio scale is the highest in the hierarchy of measurement scales. It is the
most valuable and sophisticated of all types as it has an absolute zero point along with all the
other characteristics of nominal, ordinal and interval scales. Weight, length, time interval, speed,
electrical resistance, distance and temperature in Kelvin scale are some examples of Ratio Scale
measurement. The presence of absolute or normal zero point in the Ratio Scale represents the
absence of magnitude of a variable attribute. Mathematical operations in ratio scale have equal
intervals between the numbers on the scale. Thus, If a Car A covered 50 Km in an hour in
comparison to another Car B which travelled 100 Km in an hour under similar circumstances,
then, according to the ratio scale, it is true to conclude that Car B ran twice as fast as Car A.

As pointed out earlier, a given body at 40 degrees centigrade cannot be said to possess four times
the heat or kinetic energy of heat than at 10 degrees Centigrade. On the other hand, if the
temperature of the body is measured on the Kelvin Scale, which is the true measure of kinetic
energy of heat associated with the random motion of molecules of the substance, we can say that
the body possesses four times the kinetic energy of heat or the motion of molecule is four times as
rapid at 40 degrees Kelvin as at 10 degrees Kelvin. We know that heat is internal/ kinetic energy
which an object poseeses because its molecules are in motion. At zero degree absolute Kelvin
Scale, it is a point at which matter ceases to agitate. In other words, at zero degree Kelvin,
molecules of all bodies are completely at rest or motionless. It may be pointed out that zero
degree Kelvin = -273 degree Centigrade, that is, the lowest temperature - absolute zero – which
can never be achieved.

Once the ratio scale is constructed, its numerical value can also be transformed into another unit
of the Ratio scale by multiplying each value by a constant which is expressed algeracially,

If x is the measure or value of a given attribute in some specifi units, then y = ax would be the
measured value on another scale where ‘a’ units of second kind make one unit of first kind. Thus,
length of an entity in inches (inch) can be converted to corresponding length in centimeters (cm)
by the use of the following equation:

PSYCHOLOGY PAPER 4: Applied Psychometrics

MODULE 1: Measurement and Scaling
 7. Tabular Representation of Measurement Scale Types with Major
Attributes

Table 2: Types of Measurement Scales with Major Attributes

Scale Type / Basic Empirical Mathematical Permissible Permissible Typical Example

Level Operations Relations Transformation Statistics
Possible

Nominal Determination of A = B. For example, if Y = f(x), where f(x) Number of cases Male vs Female classification; Literate vs
Equivalence or categories are male and means any one to one given set mode; Illiterate classification; Example of
Non-Equivalence female, then a Person A substitution; categories contingency psychological tests/scales using nominal
who falls in the male are mutually exclusive property is Maudsley Personality Inventory
category will not be and a person can fall in (MPI); In MPI score, test scores are
similar to another person only one category. categorized in terms of introverts and
B who falls in the extroverts.
female category in terms
of chosen attributes.

Ordinal The above plus A > B, B < A. For Y = f(x), where f(x) Median Arranging a group of students based on
determination of exampl if the IQ score means any increasing Percentile Rank their marks obtained in psychology
rank order (that is, of A is greater than B, monotonic function correlation, Sign
greater or less than) then (the only other test, Run test
pos bility) the IQ score
of B is less than A.
Hence, the mathematical
relation of symmetricity
is applicable to Ordinal
Scale.

Interval The above plus (A - B) = (C - D) Y = ax + b, where a > Mean, standard Celsius and Fahrenheit scales, Calendar
determination of 0 (two degrees of deviation, dates, IQ scales
equality of intervals freedom) average
deviation,
corelation ratio,
t-test, Product
Moment
Corelation

Ratio The above plus A/B=C/ Y = ax, where absolute Geometric Numerosity, length, weight, density,
determination of zero exists, a > 0 (one mean, Kelvin-scale, time interval, height
quality of ratio DA/B= degree of freedom) coefficient of
variation
harmonic mean,
C/D percent variation

Source: Chadha N.K., 2009

1. Scales are listed in an ascending order of power or strength. The stronger scales have the ability to perform the empirical and mathematical operations of
Notes the weaker ones. Thus, ratio Scale is supposed to perform all operations of the Interval, Ordinal and Nominal scales as well as those which are unique to
itself.
2. The scale type is defined by the manner in which numbers can be transferred without empirical information.
3. The basic operations needed to create a scale are listed in second column.
4. In Column three, lower case letters indicate numerals and upper case letters indicate numbers.
5. The fourth column gives the mathematical transformations that leave the scale from invariant. Any numeral x on a scale can be replaced by another
numeral y, which is a function of x as listed in the column.
6. The fifth column lists, cumulatively downwards, some of the statistical operations that show invariance under the transformation of the fourth column.

PSYCHOLOGY PAPER 4: Applied Psychometrics

MODULE 1: Measurement and Scaling
8. Measurement Scales in Psychology
According to Kerlinger (1973, p.439): the Ordinal level of measurement is most frequently used
in psychology. As quoted by Kerlinger:

‘Intelligence, aptitude and personality test scores are basically and strictly speaking, ordinal.
They indicate with more or less accuracy not the amount of intelligence, aptitude, and
personality traits of individuals, but rather the rank order positions of the individuals.” He
further added: “Most psychological and educational scales approximate interval equality fairly
well but if ordinal measurements were treated as if they were interval measurements, the test user
must be constantly alert to the possibility of gross inequality of intervals.” (pp. 440 – 441)

It is to be noted that most psychologists want to treat their assessment data as interval when
actually it can be best described as ordinal. This is because a greater level of flexibility and
statistical manipulations is possible with interval scale than ordinal level of measurement.
Statistical computations like mean, standard deviation, average deviation, correlation ratio, t-test,
Product Moment Correlation can be employed to interval level of measurement but not with
the ordinal scale.

9. Summary

 Different types of’ Measurement Scales’ are used to measure psychological attributes.
The appropriate statistical procedure applied to a data set depends on the measurement
scale used.

 Measurement Scales are classified depending on the type of variable being measured. A
scale used to measure a continuous variable is termed as ‘Continuous Scale’ and a scale
used to measure a discrete variable is called as a ‘Discrete Scale’.

 Measurement scales are categorized into four categories depending on the amount of
quantitative information the assigned numbers possess- Nominal Scale, Ordinal Scale,
Interval scale and Ratio Scale.

 Nominal Scale represents the lowest form of measurement in which numbers are used to
represent categories. In an Ordinal Scale, a set of objects or people is ranked in ascending
or descending order with reference to an attribute.

 Interval Scale is characterized by ranked /ordered numbers along with equivalent

difference between the numbers. A Ratio Scale is advancement of prior three categories
in which distances between numerals are stated with respect to a rational zero rather than
with respect to each other.

 The Ordinal level of measurement is most frequently used in psychology.

PSYCHOLOGY PAPER 4: Applied Psychometrics

MODULE 1: Measurement and Scaling