You are on page 1of 8

1 of 8

DATA ANALYSIS
NOTE ON SCALING TECHNIQUES

Sonam
Department of Economics,
Hansraj College,
University of Delhi

In Data Analysis, when the variables are defined for a research problem/objective, the
measurement scale that the data for the variables represent is also essential to be
defined. The measurement scale means creating a continuum and locating measured
objects on it.

Some of the questions which the measurement scale answers are:

What is the ordering of the values?


Are the differences among pairs of values equivalent?
Whether one value can be expresses in terms of another?

The four major scales used in data analysis are (Figure 1):
1. Ratio scale
2. Interval scale
3. Nominal scale
4. Ordinal scale

The type of scale used in data analysis give us an insight about the variable and is
included in the definition of the variable. The numerical/quantitative variables are
measured by either ratio or interval scale while the categorical variables are measured
by Nominal or ordinal scale.
2 of 8

Figure 1: Types of measurements scales

TYPES OF DATA

QUALITATIVE
QUANTITATIVE

MEASUREMENT

NOMINAL SCALE ORDINAL


DISCRETE CONTINUOUS

MEASUREMENT

INTERVAL SCALE RATIO SCALE

Ratio scale

The ratio scale is an ordered scale in which the differences in the measurement is a
meaningful quantity and the measurements also have a true zero point. Therefore,
points along this scale makes sense as ratios. The ratio scale enables the researcher to
compute the difference and the relative magnitudes. So for example, that variable age
or weight can be equal to zero and one can say that 50 years (or 50 kg) is twice of 25
years (or 25 kg). Time is also measured on a ratio scale because difference between
3 of 8
10 and 15 minutes is same as difference between 25 and 30 minutes and 30 minutes is
twice of 15 minutes.

EXAMPLES OF RATIO SCALE

Height (say in inches/centimetre)

Weight (say in Kg/Pounds)

Salary (In Rupees/Dollars

Years of education

Interval Scale

This scale is used for measurement of quantitative variables in a way that the
differences between the numerical labelling of the items is a meaningful quantity but
the measurements do not have a true zero point. It has all the characteristics of the
ordinal scale and it also allows us to look at the difference/distance between the
attributes. For example, measurement of temperature is a classical example of
interval scale. The difference between 15○ C and 16○ C is the same as that between
21○ C and 22○ C i.e. of 1○ C.

EXAMPLES OF INTERVAL SCALE

Temperature

Standardised SAT scores

GPA in graduation
4 of 8

FEEDBACK OF TEACHER ( CIRCLE THE APPROPRIATE NUMBER IN EACH LINE)

1.Knowledge 1 2 3 4 5

2. Clarity 1 2 3 4 5

3. Comprehensive 1 2 3 4 5

4. Doubts 1 2 3 4 5

5. Communication 1 2 3 4 5

Ordinal Scale

The scale has the feature that the objects or items are placed on the basis of their
relative position or preferences of individuals. So the data is classified into distinct
categories which implies ranking. It does tell us when there is a difference between
say “more or less” but the magnitude of difference is not know. The labelling in the
ordinal scale may be numerical but it does not hold any quantitative significance. For
example, the utility derived by a consumer from the consumption of various
consumption bundles by using an ordinal scale. In this scale, the objects are ordered
and the measurement scale reveals that bundle A is preferred over B which is in turn
preferred over C but to what extent are items preferred over one another is not
measured by the ordinal scale. Costumer satisfaction from a service can be measured
on a ordinal scale : Excellent, Good, Average, Poor, Worst and so on. In ordinal scale,
unlike interval scale and ratio scale, we do not know whether the difference between
Excellent and Good is the same as the difference between Good and Average (or
Average and Poor and Poor and worst).

EXAMPLES OF ORDINAL SCALE

FACULTY RANKING

COSTUMER SATISFACTION FROM


1. PROFESSOR
A SERVICE

2. ASSOCIATE PROFESSOR
1. EXCELLENT

2. GOOD

3. ASSISTANT PROFESSOR 3. AVERAGE

4. POOR

5. WORST
5 of 8
Nominal Scale

The nominal scale helps to classify data into distinct categories but does not imply
any ranking. It does not express any relationship between the variables and merely
act as labels. When there are only two categories, the scale is called dichotomous
scale. When we mention two categories of Yes or No, we are not saying that Yes is
better then no or vice versa or that yes is a multiple of No and we cant add/ subtract
yes and no.

Examples:

DICHOTOMOUS

OWNERSHIP OF PERSONAL GENDER

COMPUTER


1. MALE

1. YES
2. FEMALE
2. NO

QUESTION ANSWERED

1. CORRECT

2. INCORRECT

NOMINAL WITH MORE THAN TWO CATEGORIE

COLOUR OF SHOES
POLITICAL PARTY

1. BLACK
1. BJP

2. BLUE
2. AAP

3. RED
3. CONGRESS

LOCATION OF THE HOUSE

1. NORTH

2. EAST

3. SOUTH

4. WEST

6 of 8

Figure 2: Comparison of measurement scales

RATIO

CLASSIFIES DATA INTO


TEGORIES, PROVIDES RANKING,
ISTANCE IS MEANINGFUL AND
NCLUDES A TRUE ZERO POINT INTERVAL

CLASSIFIES DATA INTO CATEGORIES,


PROVIDES RANKING AND DISTANCE IS
MEANINGFUL ORDINAL

CLASSIFIES DATA INTO CATEGORIES


AND PROVIDES RANKING
NOMINAL

CLASSIFIES DATA INTO CATEGORIES

Figure 2 shows that nominal scale is least restrictive and provides the least
information on the variables in the data. But is also the simplest of all measurement
scales. Ordinal is an improvement over the nominal scale which provides ranking to
the categories. In interval scale, an additional feature is that the difference between
any pair of values of the variables is meaningful. The ratio scale is the most superior
of all as one can also compute ratios between any pair of values which will also make
sense.

Figure 3 shows that owing to the differences in the various measurement scales,
statistical operations suitable to these scale also differ. In a nominal scale, those
statistical procedures which are based on counting are possible to be calculated such
as mode and ch-square. In an ordinal scale, the statistical operations which can be
applied are limited because the size of the intervals between the categories is
unknown. Any statistical procedure that assumes equal intervals cannot be used in an
7 of 8

Figure 3: Features of scales of measurement

Characteristics Nominal Ordinal Interval Ratio


Counts Yes Yes Yes Yes

Order of the values/ No Yes Yes Yes


Ranking
Difference between pair No No
Yes Yes

of values is meaningful
Add or subtract values No No Yes Yes
of a variable
Multiply/Divide values No No
No Yes

of a variable and Ratio


calculated
True Zero point No No No Yes (Has an
Absolute zero
point)
STATISTICAL CALCULATIONS

Measures of location Mode Median, Arithmetic mean Harmonic


Mode mean/
Geometric
mean
Measures of dispersion - Semi inter- standard standard
quartile deviation, Range deviation,
range variance,
coefficient of
variation
Test of significance Chi-square Rank-order T test, F test T test, F test
correlation

ordinal scale. We cannot calculate mean but median and percentile is possible. In an
interval scale, statistical procedures which are based on addition and subtraction can
be used in addition to the procedures appropriate for nominal and ordinal scales.
Most of the common statistical procedures like Mean, Standard Deviation, Analysis
of variance, Analysis of co-variance etc. can be worked out from the data in the
interval scale. In a ratio scale, almost all statistical operations can be applied.
8 of 8

You might also like