MEASUREMENT AND SCALING

MEASUREMENT:
 It is the assignment of numbers or other symbols to characteristics of
objects according to certain pre-specified rules.
 researchers do not measure objects but some characteristics of it.
 In reality, researchers do not measure consumers but their perceptions,
beliefs, attitudes, preferences and so on.
 The idea of assigning numbers can be helpful in two ways in accurate
understanding of a phenomenon:
(1) it allows statistical testing and
(2) it helps facilitate easier communication as people have a clear idea
with regard to what 10% or 20% means worldwide.
 Scaling
 Scaling is the branch of measurement that involves the construction
of an instrument that associates qualitative constructs with
quantitative metric units.
 Scaling can be defined as the process of assigning a set of
descriptors or rules to represent the range of possible responses to a
question about a particular phenomenon.
 The appropriateness of the raw data being collected depends directly
on the scaling technique used by the researcher.
Scales of measurement: Fundamental properties
 Assignment property
 The assignment property is also referred as description or category
property.
 It refers to the researcher’s employment of unique descriptors, or
labels to identify each object within a set.
 Order property
 The order property refers to the relative magnitude between the
descriptors.
 The relative magnitude refers to three basic properties of any object
mathematically.
 If they are two objects A and B, there are three basic mathematical
possibilities:
 (1) A is greater than B; (2) A is lesser than B; and (3) A is equal to B.
 Order property helps in identifying these properties.
 Distance property
 The distance property refers to a measurement scheme where exact
difference between each of the descriptors is expressed in absolute.
 Normally, the distance property is restricted to those situations where
the raw responses represent some type of natural numerical answer.
 Origin property
 The origin property is a measurement scheme wherein exists a
unique starting point in a set of scale points.
 For the most part, the origin property refers to a numbering system
where zero is the displayed or referenced starting point in the set of
possible responses.
 Other such origin property responses could be ‘dissatisfied’, ‘neither
dissatisfied nor satisfied’, and ‘highly satisfied’.
Primary scales of measurement
 Nominal scale
 Basic scale of measurement (lowest level of measurement)
 Discrete categories
 Categorical or Dichotomous
 No natural order

 Example:
 Gender: 0=male; 1=female
 Marital status
 Colour
 Religion, etc
 The only permissible operation on the numbers in a nominal scale is
counting
 Percentages, mode, chi-square and binominal tests can be carried
out using nominal scale based data
 Ordinal scale
 Ordered categories
 Relative rankings
 Order matters but not the difference between values
 Unknown distance between rankings
 Zero is arbitrary
 Examples of ordinal scale
• ranking of sportsman
• ranking of brands
• quality rankings and organization rankings in business magazines
• socioeconomic characteristics such as occupational status
• customer satisfaction – very satisfied, satisfied, neither satisfied nor
dissatisfied, dissatisfied, very dissatisfied
• ratings
 Statistical techniques used to describe and infer information from ordinal scale
are percentile, mean, and rank-order correlation
 Interval scale
 Ordered categories – discrete or continuous
 Equal distance between values
 Possesses assignment, order and distance properties
 Zero is arbitrary
 Addition and subtraction – cannot multiply or divide
 Examples:
• Knowledge of students – zero does not mean no knowledge
• Temperature – zero degrees also has some value
 All technique applied to nominal and ordinal data can be used in
interval scale measurement
 Other statistical techniques used are
• range
• mean
• standard deviation
• t-tests
• ANOVA
• regression analysis
• factor analysis
 Ratio scale
 Most precise
 Ordered
 Exact values
 Equal intervals
 Natural zero – not arbitrary
 It contains all the four scaling properties (assignment, order, distance
and origin) and possesses the properties of nominal, ordinal, and
interval scales and in addition an origin
 Can add, subtract, multiply and divide
 Examples
• weight, height, pulse, blood pressure, time and age
 Ratio scale is used when measuring variables such as sales, cost,
customer numbers and so on
 All statistical techniques can be applied to ratio scale based data. This
includes specialised statistics such as geometric mean, harmonic
mean and coefficient of variation
SCALING TECHNIQUES
 Important characteristics for developing high quality scales
 understanding the defined problem
 establishing detailed data requirements
 identifying and developing the constructs
 understanding the complete measurement scale
 Basic strands of scaling techniques:
 Comparative scaling
 Non-comparative scaling
 Comparative scaling involves direct comparison of stimulus objects
with one another
 Non-comparative scaling involves each stimulus object being scaled
independently of the other objects in the stimulus set
Comparative scaling techniques
 Comparative scaling techniques provide a direct comparison between
stimulus objects
 The respondents are forced to choose one out of two (or many)
stimulus objects
 Researchers can identify small differences between stimulus objects
 Paired comparison scaling
 Respondents are asked to choose one among two alternatives on a
selected criterion
 The data obtained from paired comparison scaling is ordinal in nature
 When there are more than two stimuli involved, paired comparison
scaling can still be useful technique to compare various stimuli
 Example:
• Preference between brand A and brand B
• Preference between brand B and brand C
• Preference between brand A and brand C
 Rank order scaling
 It is about ranking a specific set of stimuli on a pre-defined criterion
 It is useful in understanding a specific rank order among various
stimuli
 The respondents are provided with various stimuli objects and asked
to rank the most preferred object, the second most preferred object
and so on
 Example:
• Ranking chocolates based on taste
• Ranking television brands based on picture quality
 Rank order scaling generates ordinal data and therefore lacks
distance and origin properties
 Due to the absence of distance and origin properties, rank order
scaling cannot provide an objective difference between various
stimuli objects
 Constant sum scaling
 Respondents are asked to assign a constant sum of units (could
include points, currency, and so on) to a specific set of stimulus
objects with respect to some pre-defined criterion.
 The attributes are scaled by counting the points assigned to each
criterion by all the respondents and divided by the number of
respondents.
 Using the numbers assigned researcher can easily convert constant
sum scale into rank order scale.
 Q-sort
 Q-sort can be called an extension to rank order scaling.
 It uses a rank order procedure in which objects are sorted into piles
based on similarity with respect to some pre-defined criteria.
 It provides grouping according to the respondents’ preferences
among a relative larger number of objects quickly.
Non-comparative scaling techniques
 In non-comparative scaling, researchers use whatever rating standard
seems appropriate to them
 Respondents answering non-comparative scale based questions do
not compare the object being rated either to another object or to some
specified standard
 They evaluate only one object at a time
 Continuous rating scale
 Also known as graphic rating scale in which respondents rate the
objects by placing a mark at the appropriate position on a line that
runs from one extreme criterion to the other
 The respondent is provided with the freedom here to choose a point
anywhere along the line and is not restricted to ranking only
 Itemized rating scale
 Itemized rating scales involve selection of a specific category out of
various categories pre-defined by the researcher
 A brief description is associated with each category and respondents
are asked to select the best fitting category with the stimuli object
 Likert scale
 Likert scale is one of the highly used scales in marketing research which focuses
on degree of agreement or disagreement.
 The scale is named after Rensis Likert who developed the scale.
 The respondent is presented with a series of statements about the stimulus
objects and asked to provide views on agreement or disagreement with each of
the statement.
 A typical Likert scale constitutes of five items ranging from ‘strongly disagree’ to
‘strongly agree’.
 For the ease of statistics, researchers also associated numbers with Likert scale.
 The representation of Likert scale makes it easier for the respondents
to answer the questions.
 Researchers also use variety of number systems instead of 1 to 5,
such as – 2 to + 2 or reversing the number order from 5 to 1.
 The analysis on Likert scale can be conducted on item basis or on the
basis of the total score which can be calculated for each respondent
by summing across items.
 Likert scale can also help in developing comparison constructs.
 Semantic differential scale
 Semantic differential scale includes a seven-point bi-polar scale in
comparison to Likert’s five-point scale.
 While in Likert each item number of scale is defined in semantic
differential scale the endpoints are clearly defined.
 For example, ‘satisfaction’ and ‘dissatisfaction’ can be used as the
endpoints.
 Stapel scale
 Stapel scale consists of a single criterion in the middle of an even-
numbered range of values, from -5 to +5, without a neutral point.
 The scale is generally presented vertically.
 The respondents are asked to choose a specific number describing
the stimuli object of concern on the pre-defined criterion.
SELECTING AN APPROPRIATE SCALE
 The decisions mostly pertain to
 length of scale points
 balance of the scale
 forced vs. non-forced scales
 scale description and presentation
SCALE EVALUATION
 Researchers always attempt to develop a robust and appropriate
scale to measure a specific phenomenon, but sometimes errors in
measurement can occur due to very many reasons.
 Researchers have defined errors in two broad streams
 Systematic error
 Random error
 Systematic error affects the measurement constantly while random
error, as the name suggests is random in nature.
 Errors possible
 Respondent error: respondent characteristics such as intelligence, education can
affect the test score.
 Short-term personal factors: such as fatigue, stress, anxiety
 Situational factors: such as noise in the surroundings, presence of other people
 Clarity errors: such as poor framing of question or scale
 Mechanical errors: such as poor printing, recording error and poor design
 Interviewer error: interviewer differences and their bias in interviewing
 Analysis error: inappropriate methods of analysis used
 To avoid such errors and control the research process, after
developing an appropriate scale, researcher must assess the scale
on three dynamic constructs
 Validity
 Reliability
 generalizability
 Validity
 Validity of a scale is defined as the extent to which differences in
observed scale scores reflect the true differences among objects on
the characteristics being measured.
 By testing validity researcher can decide is the scale measuring what
it is meant to measure.
 A perfectly valid scale will have no measurement errors.
 Validity can be measured by examining content, criterion and
construct validity.
 Content validity refers to the content of the scale.
 It involves a subjective but systematic evaluation of how well the
content represents the task at hand.
 Being a subjective evaluation technique it is not considered a
sufficient measure of the validity of a scale.
 Criterion validity refers to examining whether the measurement scale
performs as expected in relation to other variables selected as
meaningful criteria.

 Construct validity is the bridge between theory and the scale. It

explains the questions of what construct or characteristic the scale is
measuring and what deductions can be made concerning the theory
underlying the scale.
 Construct validity is classified into three parts:
 Convergent
 Discriminant
 Nomological validity
Strongly Disagree Somewhat Disagree Somewhat Agree Strongly Agree 1. I feel good about my work on the job.

Strongly Disagree Somewhat Disagree Somewhat Agree Strongly Agree 2. On the whole, I get along well with others at work.

3. I am proud of my ability to cope with difficulties at work.

Strongly Disagree Somewhat Disagree Somewhat Agree Strongly Agree

Strongly Disagree Somewhat Disagree Somewhat Agree Strongly Agree 4. When I feel uncomfortable at work, I know how to handle it.

Strongly Disagree Somewhat Disagree Somewhat Agree Strongly Agree 5. I can tell that other people at work are glad to have me there.

6. I know I'll be able to cope with work for as long as I want.

Strongly Disagree Somewhat Disagree Somewhat Agree Strongly Agree

Strongly Disagree Somewhat Disagree Somewhat Agree Strongly Agree 7. I am proud of my relationship with my supervisor at work.

8. I am confident that I can handle my job without constant

Strongly Disagree Somewhat Disagree Somewhat Agree Strongly Agree assistance.

Strongly Disagree Somewhat Disagree Somewhat Agree Strongly Agree 9. I feel like I make a useful contribution at work.

10. I can tell that my coworkers respect me.

Strongly Disagree Somewhat Disagree Somewhat Agree Strongly Agree
 Convergent validity focuses on how well the scale’s measurement
positively correlates with different measurements of the same scale.
 Discriminant validity refers to the fact that the scale being investigated does not
significantly correlate with other constructs that are operationalized as being
different.
 Nomological validity allows researchers to evaluate how well one particular
construct theoretically networks with other established constructs that are related
yet different.
 It is a statistical term that refers to a "nonlogical" necessity or a law of nature. It is a
form of construct validity that requires a comparison of at least two constructs and
that those constructs have a possible linkage.
 Reliability
 Reliability in research relates to consistency of results over a period of
time.
 A scale is called reliable if it produces consistent results when
repeated measurements are made.
 Systematic errors do not have an effect on reliability however random
errors do.
 There are four ways in which reliability is measured:
 Test-retest reliability
 Alternative (parallel) forms reliability
 Internal consistency reliability
 Inter-rater reliability
 In test-retest reliability measurement, same respondents are
administered identical sets of scale items at two different times
(usually 2–4 weeks).
 The degree of similarity between the measurements (measured
through correlation between both measurements) determines the
reliability.
 The higher the correlation between the two measurements, the higher
the scale reliability.
 Alternative forms reliability, relates to a measure that is obtained by
conducting assessment of the same phenomena with the participation
of the same sample group via more than one assessment method.
 Internal consistency reliability is used to assess the reliability of a
summated scale where several items are summated to form a total
score.
 each item in the scale must measure part of what the scale is
developed to measure.
 Inter-rater reliability relates to the measure of sets of results obtained
by different assessors using same methods.
 Benefits and importance of assessing inter-rater reliability can be
explained by referring to subjectivity of assessments.
 Generalizability
 Generalizability refers to the extent to which one can generalize from
the observations at hand to a universe of generalizations

 Two aspects of generalizability

 Generalizing to a population
 Generalizing to a theory