Uploaded by Priyanka Singh

General MR Pres - scales n reliability

advertisement
Measurement Process in Research Methods –
Scales and their Reliabilities
“The best way to understand something is to try and measure it”
Dr. Priyanka Singh
University of Memphis
What is research?
 Seeking systematically to add to your own body of
knowledge and that of others by the discovery of
non-trivial facts and insights




Working systematically and recording carefully
Adding to the body of knowledge - yours and others
Discovery
Non-trivial
Types of research
 Empirical
 Literature review
 Case descriptions
 Action research
Empirical
Gathers field data to support (or dismiss)
an idea or hypothesis
Literature Review
A critical and/or comprehensive review
of what others have written about a
topic.
Case Descriptions
Detailed presentation of field data gathered
from observation, interview, survey, etc.
normally from one organization
Action Research
The researcher is part of the research process
and outcome - the research is reflexive and
transparent - often undertaken as a change
programme in the researcher’s own workplace
Reliability of Latent Social-Psychological
Constructs in Management
Perspectives on Measurement in Social Sciences

What is Measurement?

Usefulness of Social Sciences Measures

Scaling of Latent Psychological Constructs with
Multiple Items
Measurement Error
The variation in the information
sought by the researcher and the
information generated by the
measurement process employed
Measurement Error Contd…
Sampling Error
 Differences that
exist between
sample value and
the true underlying
population value
Non-Sampling Error
 All the aspects of the
research process
where mistakes and
deliberate deceptions
can occur
True Score Model
A mathematical model that provides a framework for
understanding the accuracy of measurement
Xo = Xt + Xs + Xr
Xo =
Xt =
Xs =
Xr =
observed score
true score
systematic error
random error
Which leads to
Research findings
Develop a construct
Use the concept
Utilizes the scale
Which is used to
create
Identify the concept of interest
Constitutive definition
If evaluation is satisfactory,
the researcher
Evaluate the reliability
and validity of scale
Which enables the researcher
to develop an
Operational definition
That require a
researcher to
Measurement scales
Which enables the
researcher to create
Characteristics of a Good Measurement
Reliability
The outcome of the measurement
process is reproducible (Similar results
are gotten over time and across
situations
Validity
The measurement process actually
measures what it purports to measure
Sensitivity
The measurement process shows
variability in responses when it exists in
stimulus or situations
Social Sciences Vs. Number Systems
Social Sciences
Properties of the marketing
phenomenon under study vary
from situation to situation
Number System
Properties of the number system are
always the same
1 = uniqueness
2 = ordering
3 = equality of differences
4 = equality of ratio
Physical Sciences
Social Sciences
Length,
Weight
Easy
Ratio Scale
Preference,
Attitudes
Most Marketing
Phenomenon
Fall Here
Interval Scale
Ordinal Scale
Happiness,
Creativity
Very Difficult
Nominal Scale
Four Major Levels of Measurement
Level
Description
Basic
Empirical
Operations
Typical Usage
Descriptive
Statistics
Nominal
(0,1…9)
Use numerals to identify
objects, individuals,
events etc.
Determination
of identity,
categories
Classification (M/F,
Buyer/non-buyer)
Frequency,
Percentages,
Binomial test,
chi-square test
Ordinal
(5>4,
6<9)
Provide info about the
relative amount of some
characteristic posed by
an event
Determination
of order
(greater or
less)
Rankings, ratings
(preferences)
Median (mean
and variance
metric),
Percentiles,
Rank-order
correlation
Interval
(7-5 =
6-4)
Intervals between
consecutive points are
equal
Determination
of equality of
intervals
Preferred measure
of complex
concepts and
constructs
Range, Mean,
Variance,
Standard
deviation, P.M.
correlation
Ratio
(9/3 =
12/4)
Includes an absolute
zero point
Determination
of equality of
ratios
When precision
instruments are
available (sales,
age)
Mean statistics,
Coefficient of
variation
One Variable….All measurement Levels

Nominal
Are you older than 35?
( ) Yes ( ) No

Ordinal
How old are you?
( ) < 35
( ) 35-55
( ) > 55

Interval
How old are you?
( ) 25 – 34
( ) 35 – 44
( ) 45 – 54
( ) 55 – 64

Ratio
How old are you?
_________________
e.g., Likert Scale
We love our customers
SA

A

NAND

D

SD

We have an excellent customer
complaint cell
SA

A

NAND

D

SD

We believe in 1-2-1 marketing
SA

A

NAND

D

SD

In our organization products are
tailored according to customer
needs
SA

A

NAND

D

SD

SA
We are not able to anticipate customer 
behaviour
A

NAND

D

SD

Scoring a Likert Scale
0
5
1
2
-2
1
4
2
1
-1
2
3
3
0
0
3
2
4
-1
1
4
1
5
-2
2
SA
A
NAND
D
SD
Overview of Dimensionality

Concerned with homogeneity of items

Unidimensionality vs. Multidimensionality
Forms of Reliability
 Test-retest Reliability
 Internal Consistency
 Alternative Forms
Coefficient Alpha
Coefficient alpha is the basic formula for determining the
reliability based on internal consistency.

It should be applied to new measurement methods.

Coefficient alpha sets an upper limit to the reliability.

Coefficient alpha provides a good estimate of reliability in
most situations because it considers sources of errors that
are based not, strictly speaking, on the sampling of items
per se, but on the sampling of situational factors
accompanying the items.
Coefficient Alpha Contd..

The reliability of scores obtained on a sample of
items from a domain increases with the number
of items sampled. Alpha is a function of scale
length.
α = kr/1+ (k-1) r
where k is the number of items in the scale and r
is the average correlation among the items in the
scale.
Selected Readings


Psychometric Theory – Jum C. Nunnally
(1975)
Scale Development – Subhash &
Bearden (2003)
Thank You !!
Download