Transcript 3220Lecture10Chapter8

Marketing Research
4/7/2016
1

An attempt to provide an objective
estimate of a natural phenomenon
◦ e.g. measuring height
◦ or weight
Marketing Research
4/7/2016
2

Nominal Scales
◦ Categories - mutually exclusive and collectively exhaustive.

Ordinal Scales
◦ Rank order – e.g. most important to least important.

Interval Scales
◦ Ordinal + the intervals between data points are equal.

Ratio Scales
◦ Above + a meaningful absolute zero or origin.
Marketing Research
4/7/2016
3

How could you measure height using the
following levels of measurement?
◦ Nominal Scales
◦ Ordinal Scales
◦ Interval Scales
◦ Ratio Scales
Marketing Research
4/7/2016
4
Evaluate my bathroom scale:
Is it ACCURATE?
Marketing Research
4/7/2016
5

Consistent results over time.

Measures are free from random error.
Test-Retest Reliability:
Obtained by repeating the measurement
using the same instrument under as
nearly the same conditions as possible.
Marketing Research
4/7/2016
6
Testing Reliability
Test and Retest:
• One instrument is used at two different times under similar
conditions to the same people.
Equivalent Form:
• Two similar versions of an instrument are administered at
two different times to the same people.
Spilt Half:
•Two versions of the same instrument (half of the items) are
administered to different people, but compared to one another.
7
Addresses the issue of whether what we
tried to measure was actually measured.
How do I know if my bathroom scale is valid?
That is, does 160 really MEAN anything?
How could I know this with my bathroom scale?
1. JUST using the same scale
2. Using ANOTHER scale.
Marketing Research
4/7/2016
8
• Face validity
• Content validity
• Criterion-Related Validity
• Predictive Validity
• Construct Validity
• Discriminant Validity
All types of validity are related.
Marketing Research
4/7/2016
9
10
• Situation 1:
 A lack of reliability
• my scale varies a lot (150, 160, 170)
• Situation 2:
 High reliability, but lacks validity
• My scale always shows 160, but other scales are different
• Situation 3:
 Reliable, consistent, and valid
• ALL scales show a consistent 160
Marketing Research
4/7/2016
11
Scaling Defined
In order to decrease noise:
you measure something several times
• Unidimensional scaling
• Multidimensional scaling

Graphic Rating Scales
Present to respondents with a graphic continuum
typically anchored by two extremes
Marketing Research
4/7/2016
12
Honda is a good car brand:
1 Agree
2 Somewhat Agree
3 Somewhat Disagree
4.Disagree
13
Marketing Research
4/7/2016
14