Marketing Research

Ninth Edition

Chapter 12
Using Descriptive Analysis,
Performing Population
Estimates, and Testing
Hypotheses

Copyright © 2020, 2017, 2014 Pearson Education, Inc. All Rights Reserved
Learning Objectives (1 of 2)
In this chapter you will learn
12.1 What the different types of statistical analyses used in
marketing research are
12.2 What descriptive analysis is and how to do it
12.3 When to use a particular descriptive analysis measure
12.4 How to perform descriptive analyses with SPS S
12.5 How to report descriptive statistics to clients

Copyright © 2020, 2017, 2014 Pearson Education, Inc. All Rights Reserved
Learning Objectives (2 of 2)
In this chapter you will learn
12.6 The difference between sample statistics and population
parameters
12.7 How to estimate the population percentage or mean with a
confidence interval
12.8 How to obtain confidence intervals with SPSS
12.9 How to communicate confidence intervals insights to clients
12.10 What hypothesis tests are and how to perform them
12.11 How to communicate hypothesis tests insights to clients

Copyright © 2020, 2017, 2014 Pearson Education, Inc. All Rights Reserved
Where We Are
1. Establish the need for marketing research.
2. Define the problem.
3. Establish research objectives.
4. Determine research design.
5. Identify information types and sources.
6. Determine methods of accessing data.
7. Design data collection forms.
8. Determine the sample plan and size.
9. Collect data.
10. Analyze data.
11. Communicate insights.

Copyright © 2020, 2017, 2014 Pearson Education, Inc. All Rights Reserved
Types of Statistical Analyses Used in
Marketing Research

• Descriptive analysis
• Inferential analysis
• Differences analysis
• Associative analysis
• Predictive analysis

Copyright © 2020, 2017, 2014 Pearson Education, Inc. All Rights Reserved
Table 12.1 Five Types of Statistical Analyses
Used by Marketing Researchers
Type Description Example Statistical Concepts
Descriptive Analysis Summarizes basic findings Describes the typical Typical: mean, median,
(Chapter 12) for the sample respondent, describes how mode Variability: frequency
similar respondents are to distribution, range, standard
the typical respondent deviation
(variability)

Inference Analysis Determines population Estimates population values, Standard error, null
(Chapter 12) parameters, tests support or no support for hypothesis, Z value,
hypotheses hypothesized value confidence interval

Difference Analysis Determines if differences Evaluates the statistical t test of differences in

(Chapter 13) exist significance of differences in proportions or means,
the means of two groups in analysis of variance
a sample

Association Analysis Determines connections Determines if two variables Correlation, cross-tabulation

(Chapter 14) are related in a systematic
way

Relationships Analysis Finds complex relationships Determines how several Multiple regression, beta
(Chapter 15) for the variables in the independent variables are coefficient,
dataset related to a dependent R2
variable

Copyright © 2020, 2017, 2014 Pearson Education, Inc. All Rights Reserved
Descriptive Analysis
• Used by marketing researchers to describe the sample
dataset in such a way as to portray the “typical”
respondent and to reveal the general pattern of
responses.
• Descriptive measures are typically used early in the
analysis process, and become foundations for
subsequent analysis.

Copyright © 2020, 2017, 2014 Pearson Education, Inc. All Rights Reserved
Inference Analysis
• Used when marketing researchers use statistical
procedures to generalize the results of the sample to
the target population it represents.
• Such statistical procedures allow a researcher to draw
conclusions about a population based on information
contained in the dataset provided by the sample.
• Inferential statistics include hypothesis testing and
estimating true population values using confidence
intervals.

Copyright © 2020, 2017, 2014 Pearson Education, Inc. All Rights Reserved
Difference Analysis
• Used to determine the degree to which real and
generalizable differences exist in the population to
help the manager make an enlightened decision.
• Example research questions:
• Do millennials spend more money when eating at our
restaurant compared to baby boomers?
• Customers rate our ice cream 4.4/5 and and rate the
competitor’s ice cream 4.2/5. Does it actually mean
they think our product is better?

Copyright © 2020, 2017, 2014 Pearson Education, Inc. All Rights Reserved
Association Analysis
• Investigates if and how two variables are related.
• Example questions:
• Are advertising recall scores positively associated with
intentions to buy the advertised brand?
• Are expenditures on sales force training positively
associated with sales force performance?
• Depending on the statistic used, the analysis may indicate
the strength of the association and the direction of the
association between two questions on a questionnaire in
a given study.

Copyright © 2020, 2017, 2014 Pearson Education, Inc. All Rights Reserved
 Relationship Analysis
• Statistical procedures and models to help allow insight into
multiple relationships among variables.
• Statistical procedures and models are available to help the
marketing researcher understand and make forecasts about
variables of interest based on the status or behavior of other
variables.
• For example, a researcher may discover that low-income families
watch Netflix a great deal because the parents cannot afford to take
their children to movie theaters frequently.
• Regression analysis is commonly used by the marketing researcher
to understand such complex connections.
• Because marketing managers are typically looking at several factors
simultaneously, understanding connections among these factors is
very desirable for the insight it can provide into the many
relationships among the variables in a dataset.

Copyright © 2020, 2017, 2014 Pearson Education, Inc. All Rights Reserved
Disclaimer
• The intention of this course is NOT to train you to become a Statistician.
• Rather, the focus is to acquaint you with the basic concepts involved in
each of the selected measures.
• It is important for you to have a conceptual understanding of the
commonly used statistical procedures.
• More importantly, it shows you when and where each measure is
appropriately used, and to help you interpret the meaning of statistical
results once they are reported.
• We rely heavily on computer statistical program output (such as SPSS).
Understanding Data via Descriptive
Analysis

• Two sets of measures are used extensively to describe

the information obtained in a sample.
• Measures of central tendency or measures that
describe the “typical” respondent or response
• Measures of variability or measures that describe
how similar (dissimilar) respondents or responses
are to (from) “typical” respondents or responses

Copyright © 2020, 2017, 2014 Pearson Education, Inc. All Rights Reserved
Summarizing the “Typical” Respondent

• Measures of central tendency:

• Mode: a descriptive analysis measure defined as that
value in a string of numbers that occurs most often.
• Median: expresses that value whose occurrence lies in the
middle of an ordered set of values.
• Mean (or average):

Formula for a mean

n

x
Mean  x  
i
i 1

Copyright © 2020, 2017, 2014 Pearson Education, Inc. All Rights Reserved
Mean is very popular and helpful information that we can extract from data. However, it does not tell the whole
story…
 Measures of Variability: Visualizing the
Diversity of Respondents (1 of 3)
• All measures of variability are concerned with depicting
the “typical” difference between the values in a set of
values.
• There are three measures of variability:
• Frequency distribution
• Range
• Standard deviation

Copyright © 2020, 2017, 2014 Pearson Education, Inc. All Rights Reserved
Measures of Variability: Visualizing the
Diversity of Respondents (2 of 3)

• A frequency distribution is a tabulation of the number of

times that each different value appears in a particular set
of values.
• The frequency for each value divided by the total number
of observations for all values, resulting in a percent, called
a percentage distribution.

Copyright © 2020, 2017, 2014 Pearson Education, Inc. All Rights Reserved
Measures of Variability: Visualizing the
Diversity of Respondents (3 of 3)

• Range: identifies the distance between lowest value

(minimum) and the highest value (maximum) in an
ordered set of values.
• Standard deviation: indicates the degree of variation or
diversity in the values in such a way as to be translatable
into a normal or bell-shaped curve distribution.

High SD 2 samples can have exactly the same Mean, however SD will
determine the shape of the distribution bell curve.

Low SD
Copyright © 2020, 2017, 2014 Pearson Education, Inc. All Rights Reserved
Standard Deviation
Formula for a standard deviation
n

 x - x 
2

Standard deviation
i

(s)= i-1

n -1

where

n = the number of cases

xi = each individual value
x = the mean (average)
Σ signifies that all the xi values are summed

Copyright © 2020, 2017, 2014 Pearson Education, Inc. All Rights Reserved
Figure 12.1 Normal Curve

Copyright © 2020, 2017, 2014 Pearson Education, Inc. All Rights Reserved
Table 12.2 Which Descriptive Statistic to
Use When
Example Question Measurement Level Central Tendency (most Variability (Similarity of
Typical Response) Responses)
What is your gender? Nominal scale Mode Frequency and, or
percentage distribution
Rank these five brands Ordinal scale Median Cumulative percentage
from your first choice to distribution
your fifth choice

On a scale of 1 to 5, Interval scale* Mean Standard deviation and,

how does “Starbucks” or range
rate on the variety of its
coffee drinks?

About how many times Ratio scale* Mean Standard deviation and,
did you buy fast food or range
for lunch last week?

*SPS S refers to interval or ratio measurement as simply “scale”.

Copyright © 2020, 2017, 2014 Pearson Education, Inc. All Rights Reserved
 Figure 12.2

When using SPSS DESCRIPTIVES, always keep in mind the

variables being analyzed should be interval or ratio scaled.
Copyright © 2020, 2017, 2014 Pearson Education, Inc. All Rights Reserved
Figure 12.3

Copyright © 2020, 2017, 2014 Pearson Education, Inc. All Rights Reserved
Figure 12.4

Copyright © 2020, 2017, 2014 Pearson Education, Inc. All Rights Reserved
Figure 12.5

Copyright © 2020, 2017, 2014 Pearson Education, Inc. All Rights Reserved
Reporting Scale Data

Descriptive Measure For a Standard Metric Comment

Variable Table…
Average (mean) Absolutely include, as Place averages in a column
averages are the most very close to the variable
commonly used central descriptions and arrange
tendency measure for scale variables in ascending or
data descending order of the
averages

Median, mode Do not include Managers do not relate to

medians or modes of scale
data.

Standard deviation Typically include in the table If most standard deviations are
approximately equal, do not
include, as redundancy would
result.

Minimum, Maximum Include if the data have Reporting the same value
several different values. several times is redundant.

Copyright © 2020, 2017, 2014 Pearson Education, Inc. All Rights Reserved
Example Scale Variables Table

Performance of the Subshop

Feature of the Sub shop Average* Standard Deviation
Assortment of breads 4.5 0.5
Variety of subs 4.3 0.7
Variety of toppings 4.0 0.8
Freshness of bread 3.9 0.8
Freshness of toppings 3.8 0.7
Promptness of service 3.7 1.0
Cleanliness of facility 3.7 0.9
Value for the price 3.6 1.1
Generosity of toppings 3.5 1.0
Distinctive taste 3.2 1.3

*Based on a scale where 1 = “poor” and 5 = “excellent.”

Copyright © 2020, 2017, 2014 Pearson Education, Inc. All Rights Reserved
Reporting Nominal or Categorical Data
Descriptive measure For a Standard Categorical Variable Comment
Table…

Frequencies, frequency distribution Include if the researcher wants the Place frequencies in a column very
reader to note something about the close to the variable group labels (such
sample, such as a very small sample in as male, female). If appropriate,
which percentages are greatly affected arrange the categories in ascending or
by a few respondents. descending order of the percentages.
Include a total of the frequencies at the
bottom

Percents, percent distribution Absolutely include, as percentages are Place percentages in a column close to
the most commonly used descriptive the variable group labels (such as male,
measure for nominal data. female) and beside the frequencies, if
used. If appropriate, arrange the
categories in ascending or descending
order of the percentages. Include a
100% total at the bottom.

Mode Highlight, but if obvious do not report The largest percentage group is usually
in the table. readily apparent in a percent
distribution, especially if ascending or
descending order can be used.

Copyright © 2020, 2017, 2014 Pearson Education, Inc. All Rights Reserved
Sample Nominal or Categorical Variable
Table
What Time in the Day Do You Typically Visit the Subshop?

Time Period Percent

Before 12 p m 5.3%
Between 12 p m to 3 p m 56.8%
Between 3 p m to 6 p m 24.2%
After 6 p m 13.7%
Total 100.0%

Copyright © 2020, 2017, 2014 Pearson Education, Inc. All Rights Reserved
 Statistical Inference
• Whenever a probability sample is drawn from a population, it is not
sufficient to simply report the sample’s descriptive statistics, for these
measures contain a certain degree of error due to the sampling process.
• There is always some sample error that must be taken into account.
• Statistics are sample values
• Parameters are corresponding population values
• Inference is a form of logic in which you make a general statement (a
generalization) about an entire class based on what you have observed
about a small set of members of that class.
• Statistical inference is the set of procedures used in which sample
statistics are used to estimate population parameters

Copyright © 2020, 2017, 2014 Pearson Education, Inc. All Rights Reserved
Statistical Inference
• Statistical inference takes into account that large random samples are
more accurate than are small ones.
• Example:
• You surveyed 100 customers and 30 were happy with your product → We can infer
that the satisfaction level of the population is 30% ± x%
• Then once again, you surveyed 1000 customers and 350 were happy with your
product → We can infer that the satisfaction level of the population is 35% ± y%
• In that situation, with a larger sample, you would be more certain that the sample
statistic was accurate as an estimate of the true population value y% < x%
Two Types of Statistical Inference
• A parameter estimate is used to approximate the
population value (parameter) through the use of
confidence intervals.
• For an example of parameter estimates, suppose you work for a restaurant that is
studying their customer satisfaction. It would be too costly to measure every
customer you served. Instead, you randomly sample 100 customers and measure
the satisfaction on the scale of 10. The mean of the sample is 9.2. This is the point
estimate for the population mean (μ). You also create a 95% confidence interval
for μ which is (8.8, 9.6). This means that you can be 95% confident that the true
value of the customer satisfaction for all the customer is between 8.8 and 9.6.

• Hypothesis testing is used to compare the sample

statistic with what is believed (hypothesized) to be the
population value prior to undertaking the study.
• For example, your restaurant manager collects the sample of 100 random
customers and check their bills (maybe some higher than $50, some lower than
$50). He then has the impression that that the mean amount of money all
customers spend on their meal at your restaurant is more than $50.
Therefore, he was interested in testing the hypotheses:
H : μ = 50
0

H : μ > 50
A
Copyright © 2020, 2017, 2014 Pearson Education, Inc. All Rights Reserved
Parameter Estimation: Estimating the
Population Percent or Mean
• Parameter estimation is the process of using sample
information to compute an interval that describes the
range of a parameter such as the population mean or the
population percentage.
• It involves the use of three values:
• The sample statistic (such as the mean or the
percentage)
• The standard error of the statistic
• The desired level of confidence (usually 95% or
99%)

Copyright © 2020, 2017, 2014 Pearson Education, Inc. All Rights Reserved
Statistical Inference: Sample Statistics and
Population Parameters
• A sample statistic is usually a mean or percentage.
• Examples: Average number of times eating out/month, average amount of money
spent on each meal.
Percentage of students in a high school choosing Taco Bell for their lunch.

• Standard error is the measure of variability in the

sampling distribution. (Learnt in Chapter 10)

• A confidence interval is the degree of accuracy desired

by the researcher stated in the form of a range with an
upper and lower boundary.
Copyright © 2020, 2017, 2014 Pearson Education, Inc. All Rights Reserved
Figure 12.6

Higher variability
creates higher standard
error.

Copyright © 2020, 2017, 2014 Pearson Education, Inc. All Rights Reserved
Confidence Interval
Because we only know the mean/percentage of the sample we collected the
data from. When inferring that result to the whole population, we want to
create a range with lowest and highest value that the “true”
mean/percentage of the population is supposed to fall into, with 90%, 95%,
or 99% certainty.

Formula:

Copyright © 2020, 2017, 2014 Pearson Education, Inc. All Rights Reserved
Table 12.3 How to Compute a Confidence
Interval for a Mean or a Percentage
Step 1. Find the sample statistic, either the mean, x, or the percentage, p.
Step 2. Identify the sample size, n.
Step 3. Determine the amount of variability found in the sample in the form of
standard error of the mean, Sx ,
S
Sx 
n
or standard error of the percentage, sp.
pq
Sp 
n
Step 4. Decide on the desired level of confidence to determine the value for
z : z.95 1.96  or z.99 2.58 .
Step 5. Compute your (95%) confidence interval as x 1.96 Sx or p  1.96Sp .

Copyright © 2020, 2017, 2014 Pearson Education, Inc. All Rights Reserved
 Figure 12.7
Auto Concepts Dataset:

From descriptive analysis, we found that

the average disagree–agree response to
the statement “Gasoline emissions
contribute to global warming” was 4.6,
or “agree.”

To determine the 95% confidence

interval for this average, we use “One-
Sample T Test”

Copyright © 2020, 2017, 2014 Pearson Education, Inc. All Rights Reserved
 Figure 12.8

The average is 4.62, and the 95%

confidence interval is 4.51–4.72.
Although a “5” is the code for
“Agree,” this confidence interval is
sufficiently close that we can claim it
amounts to “Agree.”

Our interpretation of this finding is

that, if we conducted a great many
replications of this survey using the
same sample size, we would find
that 95% of these sample averages
fall between 4.51 and 4.72.

Copyright © 2020, 2017, 2014 Pearson Education, Inc. All Rights Reserved
Hypothesis Tests
• In some cases, the marketing researcher or marketing
manager may state a belief about the population
parameter (either the mean or the percentage) based on
prior knowledge, assumptions, or intuition.
• This belief, called a hypothesis, most commonly takes
the form of an exact specification as to what the
population parameter value is.
• Tests of a hypothesized population parameter value:
• Test of a hypothesis about a percent
• Test of a hypothesis about a mean
• The crux of statistical hypothesis testing is the sampling
distribution concept.

Copyright © 2020, 2017, 2014 Pearson Education, Inc. All Rights Reserved
Sampling Distribution Concept
• The crux of statistical hypothesis testing is the sampling distribution concept:
• Our actual sample is one of the many theoretical samples creating the bell-shaped
curve of possible sample results, using the hypothesized value as the center of the
distribution.
• If the person who stated the hypothesis is correct, there is a greater probability of
finding a sample result close to the hypothesized number than of finding one that is far
away.
• So, if our sample value turns out to be within ±1.96/±2.58 standard errors of the
hypothesized mean, it supports the hypothesis maker at the 95%/99% level of
confidence, because it falls within 95%/99% of the area under the curve.
Hypothesis Test About a Percent
Formula for test of a hypothesis about a percent

p H
Z
Sx

where

p = sample percentage
πH = hypothesized population percentage
Sp = standard error of the percentage
If the computed z value is greater than 1.96 or less than −1.96,
we are 95% confident that the sample evidence does not
support the hypothesized parameter value.
If the computed z value is greater than 2.58 or less than −2.58,
we are 99% confident that the sample evidence does not
support the hypothesized parameter value.© 2020, 2017, 2014 Pearson Education, Inc. All Rights Reserved
Copyright
Example
• Marketing manager thinks that only 10% of customers open advertising emails
sent by your company. However, when collecting the survey from 1000 random
customers, 80% of them say they do open your company advertising emails.
Calculation of a test of the manager’s hypothesis that
only 10% of customers open your advertising emails.

Interpretation: Marketing manager was wrong because his hypothesis of 10% of

customers opening advertising emails was 55.3 (which is >1.96) standard errors
away from the 80% finding of the sample.
Hypothesis Test About a Mean
Formula for test of a hypothesis about a mean
x  H
Z
Sx

where

x  sample mean
H  hypothesized population mean
Sx  standard error of the mean
If the computed z value is greater than 1.96 or smaller than
−1.96, we are 95% confident that the sample evidence does not
support the hypothesized parameter value.
If the computed z value is greater than 2.58 or smaller than
−2.58, we are 99% confident that the sample evidence does not
support the hypothesized parameter value.
Copyright © 2020, 2017, 2014 Pearson Education, Inc. All Rights Reserved
Example
• I believe that a household in Sydney spend about $2750 a month on
living expenses on average. I collected a sample of 100 random
households and got a mean of $2800.

Calculation of a test for my hypothesis that Sydney

households spend an average of $2800 a month.

Interpretation: I was correct because my hypothesis of average spending $2750 was 1.43 (which is <1.96) standard
errors away from the $2800 finding of the sample.
Figure 12.9

Copyright © 2020, 2017, 2014 Pearson Education, Inc. All Rights Reserved
Table 12.4 The Five Basic Steps Involved in
Hypothesis Testing
Step 1. Begin with a statement about what you believe to be the population
mean( H ) or percentage ( H ).
Step 2. Draw a random sample and determine the sample mean (x )
or percent (p).
Step 3. Compare the statistic to the hypothesized parameter, then divide by the
standard error to compute z.
x  H p H
Z Z
Sx Sp

Step 4. If z is within  1.96 /  2.58 standard errors, it supports the hypothesis

at the 95%/99% level of confidence. (Alternatively, the exact degree of support
can be assessed on SPSS output.)
Step 5. If the sample does not support the hypothesis, revise the hypothesis to be
consistent with the sample’s statistic using the confidence interval formula.

Copyright © 2020, 2017, 2014 Pearson Education, Inc. All Rights Reserved
 Figure 12.10
In Auto Concepts Dataset:
• On our scale the “neutral” position
corresponds to the value code of
“4.”
• We will hypothesize that the
general public is neutral to the
statement “I am worried about
global warming.”
• To perform a mean hypothesis test,
SPSS provides a Test Value box in
which the hypothesized mean can
be entered.
• Use the ANALYZE-COMPARE
MEANS-ONE SAMPLE T TEST
command sequence.
• Next, enter “4” as the Test Value
and click the OK button.

Copyright © 2020, 2017, 2014 Pearson Education, Inc. All Rights Reserved
Figure 12.11

Copyright © 2020, 2017, 2014 Pearson Education, Inc. All Rights Reserved