Case 12.2 Integrated Case: Global Motors Descriptive and Inference Analysis

350



Chapter 12 • Using DesCriptive analysis, performing popUlation estimates, anD testing hypotheses



Question Description



Codes



Value Labels



Size of home town or city



1,2,3,4,5



Gender

Marital status

Number of people in family

Age category

Education category



0,1

0,1

Actual number

1,2,3,4,5

1,2,3,4,5



Income category



1,2,3,4,5



Dwelling type



1,2,3,4



I am worried about global warming.

Gasoline emissions contribute to global

warming.

Desirability: 1-seat all-electric model

Desirability: 4-seat all-electric model

Desirability: 4-seat gasoline hybrid model

Desirability: 4-seat diesel hybrid model

Desirability: 5-seat standard size gasoline

model

Lifestyle: Novelist

Lifestyle: Innovator

Lifestyle: Trendsetter

Lifestyle: Forerunner

Lifestyle: Mainstreamer

Lifestyle: Classic

Favorite television show type



1,2,3,4,5,6,7



1 million and more, 500K to 1 million,10K to

500K,10K to 100K,Under 10K

Male, Female

Unmarried, Married

No labels

18 to 24, 25 to 34, 35 to 49, 50 to 64, 65 and older

Less than high school, high school diploma, some

college, college degree, postgraduate degree

Under $25K, $25K to 49K, $50K to 74K, $75K to

125K, $125K and more

Single family, Multiple family, Condominium/

townhouse, Mobile home

Very strongly disagree, Strongly disagree, Disagree,

Neither disagree nor agree, Agree, Strongly Agree,

Very strongly agree

Very undesirable, Undesirable, Somewhat desirable, Neutral, Somewhat desirable, Desirable, Very

desirable



Favorite radio genre



1,2,3,4,5,6



Favorite magazine type



1,2,3,4,5,6,7,8



Favorite local newspaper section



1,2,3,4,5,6,7



Use of online blogs

Use of content communities

Use of social network sites

Use of online games

Use of virtual worlds



1,2,3,4



1,2,3,4,5,6,7



1, … ,7



Does not describe me at all, …, Describes me

perfectly



1,2,3,4,5,6,7



Comedy, Drama, Movies/mini-series, News/documentary, Reality, Science fiction, Sports

Classic pop & rock, Country, Easy listening, Jazz &

blues, Pop & chart, Talk

Business & money, Music & entertainment, Family

& parenting, Sports & outdoors, Home & garden,

Cooking, food, & wine, Trucks, cars, & motorcycles, News, politics, & current events

Editorial, Business, Local news, National news,

Sports, Entertainment, Do not read

Never, …, 4+ times per day



revieW QUestions/appliCations



For each question below, it is your task to determine

the type of scale for each variable, conduct the proper descriptive analysis with SPSS, and interpret it.

1. What is the demographic composition of the sample?

2. How do respondents feel about: (1) global warming and

(2) gasoline emissions?

3. What type of automobile model is the most desirable to

people in the sample? What type is the least desirable?

4. Describe the “traditional” media usage of respondents

in the sample.

5. Describe the social media usage of the respondents in

the sample.

6. The Global Motors principals believe that the desirability on the part of the American public for each



351



of the automobile models under consideration is the

following.

Vehicle Model Type

1-seat all-electric

4-seat all-electric

4-seat gasoline hybrid

5-seat diesel hybrid

5-seat standard size gasoline



Desirability*

3

4

4

3

2



*Measured on 1-7 scale.



Test these hypotheses with the findings from the

survey.



CHAPTER



13

Learning Objectives

• Tolearnhowdifferencesare

usedformarketsegmentation

decisions



The Importance of Differences Analysis

Marketers often want to know if specific strat-



• Tounderstandwhen t testsor

ztestsareappropriateandwhy

youdonotneedtoworryabout

thisissuewhenusingSPSS



egies or tactics had an impact in the market-



• Tobeabletotestthedifferences

betweentwopercentagesor

meansfortwoindependent

groups



might want to know if consumers changed



place in terms of “moving the needle” on key

performance indicators. For example, they

their opinions about a brand (and their willingness to consider purchasing it) based on their



• Toknowwhatapairedsamples

differencetestisandwhento

useit



Kartik Pashupati,

Research Manager,

Research Now®



• TocomprehendANOVAandhow

tointerpretANOVAoutput



set of questions) both before and after exposure and then examine the



• Tolearnhowtoperform

differencestestsformeans

usingSPSS



example, on an intentions to purchase scale, the before average might



“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.







Implementing Basic

Differences Tests



10 Analyze data.

11 Prepare and present the final



research report.



exposure to an ad campaign. This means market researchers need to measure the same indicators (i.e., ask the target audience the same



differences in scores to see if the ad campaign made a difference. For

be a 3, which moves to a 4 after the ad campaign has run.

Some marketers are comfortable about using mean scores to examine changes in key performance indicators, but many others prefer

to examine changes in proportions, such as changes in the percentage

of survey respondents who checked the “top two boxes” in a battery

of questions. (Marketers use the term “top two boxes” as a shorthand

way of referring to the percentage of respondents who selected the two

most positively worded items in a Likert-type question. For example, in a

customer satisfaction survey, the top two boxes refer to the percentage

of customers who said they were “Satisfied” or “Very Satisfied” with a

product or service.) While mean scores allow for much more powerful

statistical tests, they can sometimes suppress interesting results if many

responses cluster around the neutral or midpoint of a scale. In such

instances, marketers often find differences in proportions much more

intuitive and useful for decision making, as they are most interested in

making a difference among those customers who have strong positive

(or negative!) opinions.



Even if the data indicate that the marketing strategy moved the needle

in the desired direction, researchers need to use statistical tests to ensure that

the observed changes can indeed be attributed to the marketing strategy and

to rule out the possibility that the differences are due to random or chance

factors.



Text and Images: By permission,

Research Now®



K



eeping Mr. Pashupati’s comments in mind, in this chapter, we describe the logic of

differences tests, and we show you how to use SPSS to conduct various types of differ ences tests.1 We begin this chapter discussing why differences are important to marketing managers. Next, we introduce differences (percentages or means) between two independent

groups, such as a comparison of high-speed cable versus DSL telephone Internet users on how

satisfied they are with their Internet connection service. Next, we introduce ANOVA, a scary

name for a simple way to compare the means of several groups simultaneously and to quickly

spot patterns of significant differences. We provide numerical examples and share examples

of SPSS procedures and output using the Global Motors survey data. Finally, we establish that

it is possible to test for a difference between the averages of two similarly scaled questions.

For instance, do buyers rate a store higher in “merchandise selection” than they rate its “good

values”?



Why Differences Are Important

Perhaps one of the most useful marketing management concepts is market segmentation. Basically, market segmentation holds that different types of consumers have different requirements, and these differences can be the bases of marketing strategies. As an example, Iams,

which markets pet foods, has more than 20 varieties of dry dog food geared to the dog’s

age (puppy, adult, senior), weight situation (small, medium, large), and activity (reduced,

normal, moderate, high). Toyota Motors

has 20 models, including 8 cars, 2 trucks,

7 SUVs and vans, and 3 hybrids. Even

Boeing has six types of commercial

jets, including the newly developed

Dreamliner, plus a separate business jets

division for corporate travel.

Let’s look at differences from the

consumer’s side. Everyone washes his or

her hands, but the kind of soap required

differs for weekend gardeners with potting soil under their fingernails, factory

workers whose hands are dirty with

solvents, preschoolers who have sticky

drink residue on their hands and faces,

or aspiring beauty princesses who wish

their hands to look absolutely flawless.

The needs and requirements of each of

Photo: © NAN/Fotolia

these market segments differ greatly



353



354



Chapter 13 • ImplementIng BasIC DIfferenCes tests



Market segmentation

is based on differences

between groups of

consumers.



To be potentially useful to

the marketing researcher

or manager, differences

must, at minimum, be

statistically significant.



To be useful to the

marketing researcher or

manager, differences must,

if statistically significant,

be meaningful.



To be useful to the

marketing researcher or

manager, differences must,

if statistically significant

and meaningful, be stable.



from the others, and an astute marketer will customize his or her marketing mix to each target market’s unique situation.2

These differences may seem quite obvious, but as competition intensifies, prolific market segmentation and target marketing become the watchwords of most companies in an

industry. Consumer marketers need to investigate differences among consumer groups, and

B2B marketers seek to differentiate the needs and preferences among business establishments. One commonly used basis for market segmentation is the discovery of (1) statistically

significant, (2) meaningful, (3) stable, and (4) actionable differences. We will discuss each

requirement briefly. In this discussion, we will use the example of working with a pharmaceuticals company that markets cold remedies.

The differences must be significant. As you know, the notion of statistical significance underpins marketing research.3 Statistical significance of differences means that

the differences found in the sample(s) truly exist in the population(s) from which the

random samples are drawn. Apparent differences between and among market segments

must be subjected to tests that assess their statistical significance. This testing is the topic

of this chapter, and we will endeavor to teach you how to perform these tests and interpret

the results.

For example, we could ask cold sufferers, “How important is it that your cold remedy

relieves your (insert symptom here)…” The respondents would respond using an scale of

1 = not important to 10 = very important for each cold symptom such as fever, sore throat,

congestion, aching muscles, etc., and statistical tests such those described in this chapter

would determine if the responses were significantly different. With those in the grip of a

cold virus, we might find two groups that have statistically significant differences. Congestion sufferers greatly desire breathing congestion relief, whereas muscle aches and pains

sufferers more greatly desire relief from musculoskeletal aches and pains associated with

their colds.

The differences must be meaningful. A finding of statistical significance in no way

guarantees “meaningful” difference. In fact, with the proliferation of data mining analysis

due to scanner data with tens of thousands of records, online surveys that garner thousands

of respondents, and other ways to capture very large samples, there is a real danger of finding a great deal of statistical significance that is not meaningful. The reason for this danger

is that statistical significance is determined to a high degree by sample size.4 You will see

in this chapter by examining the formulas we provide that the sample size, n, is instrumental in the calculation of z, the determinant of the significance level. Large samples—those

in excess of 1,000 per sample group—often yield statistically significant results when the

absolute differences between the groups are quite small. A meaningful difference is one

that the marketing manager can potentially use as a basis for marketing decisions.

In our common cold example, there are some meaningful implications that one group

cannot breathe easily while the other group has aches and pains as there are cold remedy ingredients that reduce congestion and other ingredients that diminish pain. Granted, the pharmaceuticals company could include both ingredients, but the congestion sufferers do not want

an ingredient that might make them drowsy due to the strong pain relief ingredient, and the

aches and pains sufferers do not want their throats and nasal passages to feel dry and uncomfortable due to the decongestant ingredient. These differences are meaningful both to the

customer groups and to the pharmaceuticals manufacturer.

The differences should be stable. Stability refers to the requirement that we are not

working with a short-term or transitory set of differences. Thus, a stable difference is one

that will be in place for the foreseeable future. The persistent problem experienced by congestion sufferers is most probably due to some respiratory weakness or condition. They may

have preconditions such as allergies or breathing problems, or they may be exposed to heavy

pollution or some other factor that affects their respiration in general. Muscle aches and

pains sufferers may be very active people who do not have respiration weaknesses, but they



small sample sIzes: the Use of a t test or a z test anD how spss elImInates the worry



355



may value active lifestyle practices, such as regular exercise, or their occupations may require a good deal of physical activity. In either case, there is

a good possibility that when a cold strikes, the sufferer will experience the

same discomfort, either congestion or muscle aches, time and time again.

That is, the differences between the two groups are stable. The pharmaceuticals company can develop custom-designed versions of its cold relief

product because managers know from experience and research that certain

consumers will be consistent (stable) in seeking certain types of relief or

specific product benefits when they suffer from colds.

The differences must actionable. Market segmentation requires that

standard or innovative market segmentation bases are used and that these

bases uniquely identify the various groups so that they can be analyzed

and put in the marketer’s targeting mechanisms. An actionable difference

means that the marketer can focus various marketing strategies and tactics, such as product design or advertising, on the market segments to

accentuate the differences between the segments. A great many segmentation bases are actionable, such as demographics, lifestyles, and product

benefits. In our example, among the many symptoms manifest by colds

sufferers, we have identified two meaningful and stable groups, so a cold

remedy product line that concentrates on each one of these separately is

possible. A quick glance at the cold remedies section of your local drug

Because cold suffers consistently have

store will verify the actionability of these cold symptoms market segments.

different symptoms, such as runny

You may be confused about meaningful and actionable differences. Renoses, congestion, and achy muscles,

call that we used the words “potentially use” in our definition of a meaningful

pharmaceutical companies have identified

difference. With our cold remedies example, a pharmaceutical company could

different market segments.

potentially develop and market a cold remedy that is specific to every type of

cold symptom as experienced by every demographic group and further identiPhoto: Andrii Oleksiienko/Shutterstock

fied by lifestyle differences. For example, there could be a cold medicine to

alleviate the runny noses of teenage girls who participate in high school athletics and a different To be useful to the

one for the sniffles in teenage boys who play high school sports. But it would be economically manager, differences must

unjustifiable to offer so many different cold medicines, so all marketers must assess actionabil- be statistically significant,

meaningful, stable and

ity based on market segment size and profitability considerations. Nevertheless, the fundamen- actionable.

tal differences are based on statistical significance, meaningfulness, and stability assessments.

To be sure, the bulk of this chapter deals strictly with statistically significant differences,

because it is the beginning point for market segmentation and savvy target marketing. Meaningfulness, stability, and actionability are not statistical issues; rather, they are marketing

manager judgment calls.



Small Sample Sizes: The Use of a t Test or a z Test

and How SpSS Eliminates the Worry

Most of the equations related in this chapter will lead to the computation of a z value. As

we pointed out in the previous chapter, computation of the z value makes the assumption

that the raw data for most statistics under scrutiny have normal or bell-shaped distributions. However, statisticians have shown that this normal curve property does not occur

when the sample size is 30 observations or less.5 In this instance, a t value is computed

instead of a z value. The t test is defined as the statistical inference test to be used with

small samples sizes (n # 30). Any instance when the sample size is 30 or greater requires

the use of a z test.

The great advantage to using statistical analysis routines on a computer is that they

are programmed to compute the correct statistic. In other words, you do not need to decide whether you want the program to compute a t value, a z value, or some other value.



®

The t test should be used

when the sample size is

30 or less.



356



Chapter 13 • ImplementIng BasIC DIfferenCes tests



Most computer statistical

programs report only

the t value because it is

identical to the z value

with large samples.



With SPSS, the analyses of differences are referred to as t tests, but now that you realize that

SPSS will always determine the correct significance level, whether it is a t or a z, you do not

need to worry about which statistic to use. The talent you need to acquire is how to interpret

the significance level reported by SPSS. Marketing Research Insight 13.1 introduces a “flag

waving” analogy that students have told us is helpful in this regard.



Marketing research insight 13.1



Practical Application



Green Flag Signals and Significance in Statistical Analysis

The output from statistical procedures in all software programs

can be envisioned as “green flag” devices. When the green signal flag is waving, statistical significance is present. Then, and

only then, is it warranted to look at the findings more closely to

determine the pattern of the findings; if the flag is not green,

your time will be wasted by looking any further. To read statistical flags, you need to know two things. First, where is the flag

located? Second, what color is it?



The flags, called “p values” by statisticians, are identified on

computer output by the terms significance or probability.

Sometimes abbreviations such as “Sig” or “Prob” are used to

economize on the output. To find the flag, locate the “Sig” or

“Prob” designation in the analysis, and look at the number associated with it. The number will be a decimal, perhaps as low

as 0.000 but ranging to as high as 1.000. When you locate it,

you have found the statistical significance flag.



Where Is the Flag?



What Color Is the Flag?



Virtually every statistical test or procedure involves the computation of some critical statistic, and that statistic is used to

determine the statistical significance of the findings. The critical statistic’s name changes depending on the procedure and

its underlying assumptions, but usually the statistic is identified

as a letter, as in z, t, or F. Statistical analysis computer programs

automatically identify and compute the correct statistic, so although it is helpful to know ahead of time what statistic will be

computed, it is not essential to know it. Moreover, the statistic is

not the flag; rather it is just a computation necessary to raise the

flag. You might think of the computed statistic as the flagpole.

The computer program also raises the flag on the flagpole, but its name changes a bit depending on the procedure.



In NASCAR racing, the green flag signals the start of the race.

For purposes of this textbook, we have adopted the 95% level

of confidence. That is, if you are 95% confident that the green

flag is out, you would expect the race to be under way.

As we noted previously, the significance or probability values reported in statistical analysis output range from .0000 to

1.000, and they indicate the degree of support for the null hypothesis (no differences). If you take 1 minus the reported significance level—for example, if the sig level is .03, you would

take 1 minus .03 to arrive at .97, or 97%—that is the level of

confidence for our finding. Any time this value is .05 or less

(95% level of confidence), you should know you have the

green flag to start your interpretation of the findings.



Testing for Significant Differences

Between Two Groups

Statistical tests are used

when a researcher wants

to compare the means

or percentages of two

different groups or

samples.



Independent samples are

treated as representing

two potentially different

populations.



Often, as we have done in our cold remedy example, a researcher will want to compare two

groups of interest. That is, the researcher may have two independent groups such as first-time

versus repeat customers, and he or she may want to compare their answers to the same question. The question may be either a nominal or an ordinal scale. Such a variable requires that the

researcher compare percentages; a scale variable requires comparing means. As you know by

now, the formulas differ depending on whether percentages or means are being tested.

Differences betWeen Percentages With tWO grOuPs

(inDePenDent saMPLes)

When a marketing researcher is interested in making comparisons between two groups of

respondents to determine whether there are statistically significant differences between

them, in concept, he or she is considering them as two potentially different populations.



testIng for sIgnIfICant DIfferenCes Between two groUps



The question to be answered becomes whether their respective population parameters are different. (A parameter is simply a value in the population that is of interest to the researcher.)

But, as always, a researcher can only work with the sample results. Therefore, the researcher

must fall back on statistical significance to determine whether the difference that is found between the two sample statistics is a true population difference. You will shortly discover that

the logic of differences tests is similar to the logic of hypothesis testing, which was discussed

in the previous chapter.

To begin, we will refer to an intuitive approach you use every day when comparing two

things to make an inference. Let’s assume you have read a Business Week article about college

recruiters that quotes a Harris poll of 100 randomly selected companies, indicating that 65%

will be visiting college campuses to interview business majors. The article goes on to say that

a similar poll taken last year with 300 companies found that only 40% would be recruiting at

college campuses. This is great news: More companies will be coming to your campus this

year with job interviews. However, you cannot be completely confident of your joyous conclusion because of sampling error. If the difference between the percentages was very large,

say 80% for this year and 20% for last year, you would be more inclined to believe that a true

change had occurred. But if you found out the difference was based on small sample sizes,

you would be less confident with your inference that last year’s and this year’s college recruiting are different. Intuitively, you have taken into account two critical factors in determining

whether statistically significant differences exist between a percentage or a mean compared

between two samples: the magnitude of the difference between the compared statistic (65%

versus 40%) and sample sizes (100 versus 300).

To test whether a true difference exists between two group percentages, we test the null

hypothesis, or the hypothesis that the difference in their population parameters is equal

to zero. The alternative hypothesis is that there is a true difference between them. To perform the test of significance of differences between two percentages, each representing

a separate group (sample), the first step requires a comparison of the two percentages. The

comparison is performed to find the arithmetic difference between them. The second step

requires that this difference be translated into a number of standard errors away from the

hypothesized value of zero. Once the number of standard errors is known, knowledge of

the area under the normal curve will yield an assessment of the probability of support for

the null hypothesis.

For a difference between two percentages test, the equation is as follows:

Formula for significance of the difference

between two percentages



z =



357



With a differences test,

the null hypothesis states

that there is no difference

between the percentages

(or means) being

compared.



p1 - p2

sp 1 - p 2



Where

p1 = percentage found in sample 1

p2 = percentage found in sample 2

sp1 - p2 = standard error of the difference between two percentages

The standard error of the difference between two percentages combines the standard error of the percentage for both samples, and it is calculated with the following formula:

Formula for the standard error of the

difference between two percentages



sp 1 - p 2 =



A



p2 * q2

p1 * q1

+

n1

n2



Again, if you compare these formulas to the ones we used in hypothesis testing in

Chapter 12, you will see that the logic is identical. First, in the numerator, we subtract

one sample’s statistic (p2) from the other sample’s statistic (p1) just as we would subtract



With a differences test, you

test the null hypothesis

that no differences exist

between the two group

means (or percentages).



358



Chapter 13 • ImplementIng BasIC DIfferenCes tests



the hypothesized percent from the sample percent in hypotheses testing. We use the

subscripts 1 and 2 to refer to the two different sample statistics. Second, the sampling

distribution is expressed in the denominator. However, the sampling distribution under

consideration now is the assumed sampling distribution of the differences between the

percentage rather than the simple standard error of a percentage used in hypothesis testing. That is, the assumption has been made that the differences have been computed for

comparisons of the two sample statistics for many repeated samplings. If the null hypothesis is true, this distribution of differences follows the normal curve with a mean equal

to zero and a standard error equal to one. Stated somewhat differently, the procedure

requires us to accept the (null) hypothesis as true until it lacks support from the statistical

test. Consequently, the differences of a multitude of comparisons of the two sample percentages generated from many, many samplings would average zero. In other words, our

sampling distribution is now the distribution of the difference between one sample and

the other, taken over many, many times.6 The following example will walk you through

the point we just made.

Here is how you would perform the calculations for the Harris poll on companies coming to campus to hire college seniors. Recall that last year’s poll with 300 companies reported

that 40% were coming to campus, while this year’s poll with 100 companies reported that

65% were visiting campuses.

Computation of the

significance of the

difference between

two percentages



z =

=



=



p1 - p2

sp 1 - p 2

65 - 40

65 * 35

40 * 60

+

A 100

300

25



222.75 + 8.0

25

=

5.55

= 4.51



When z is greater than

1.96, there is little support

for the null hypothesis of

no difference. Therefore,

there is a statistically

significant difference.



Notes:

p1 = 65%

p2 = 40%

n1 = 100

n2 = 300



We compare the computed z value with our standard z of 1.96 for the 95% level of confidence, and the computed z of 4.51 is larger than 1.96. A computed z value that is larger than

the standard z value of 1.96 amounts to no support for the null hypothesis at the 95% level

of confidence. There is a statistically significant difference between the two percentages, and

we are confident that if we repeated this comparison many, many times with a multitude of

independent samples, we would conclude that there is a significant difference in at least 95%

of these replications. Of course, we would never do many, many replications, but this is the

statistician’s basis for the level of significance.

We realize that it is confusing to keep in mind the null hypothesis, to understand all the

equations, and to figure out how to interpret the findings. We have provided a table that describes the null hypothesis for each type of group differences test described in this chapter.

Refer to Table 13.1.

It is a simple matter to apply the formulas to percentages to determine the significance of

their differences, for all that is needed is the sample size of each group. Marketing Research

Insight 13.2 relies on significance of the difference between percentages tests we have computed based on the information in the source. This feature highlights the significantly different

profiles of grocery item impulse purchasing found for French versus Swedish supermarket

customers.



testIng for sIgnIfICant DIfferenCes Between two groUps



Table 13.1



359



Null Hypotheses for Group Differences Tests



null hypothesis



what Does It mean if the hypothesis Is

not supported?



Differences between two group percents

No difference exists between the percents of the

two groups (populations).



A difference does exist between the percents of

the two groups (populations).



Differences between two group means

No difference exists between the means of the

two groups (populations).



A difference does exist between the means of

the two groups (populations).



Differences in means among more than two groups (Note: Only differences in means

can be tested here)

No difference exists between the means of all

A difference exists between the means of at

paired groups (populations).

least one pair of groups (populations).



Active Learning

Calculations to Determine Significant Differences

Between percentages

You can now perform your own tests of the differences between two percentages using the

formulas we have provided and described. A local health club has just finished a media blitz

(newspaper, television, radio, etc.) for new memberships. Whenever prospective members

visited the health club’s facilities, they were asked to fill out a short questionnaire, and one

question asked them to indicate what ads they saw in the past month. Some of these prospects joined the health club, while some did not; thus, we have two populations: those who

joined the health club and those who did not. At the end of the 30-day campaign, a staff

member performed the following tabulations.



 

Total visitors



Joined the Health Club



Did Not Join the Health Club



100



30



Recalled newspaper ads



45



15



Recalled FM radio station ads



89



20



Recalled Yellow Pages ads



16



5



Recalled local TV news ads



21



6



Use your knowledge of the formula and the test of the significance of the difference between

two percentages to ascertain if there are any significant differences in this data. What are the

implications of your findings with respect to the effectiveness of the various advertising media

used during the membership recruitment ad blitz?



To learn

about

proportion

differences

tests,

launch 

www.youtube.com, and

search for “Hypothesis

Test Comparing Population

Proportions.”



®



using sPss fOr Differences betWeen Percentages Of tWO grOuPs

As is the case with most statistical analysis programs, SPSS does not perform tests of the

significance of the difference between the percentages of two groups. You can, however, use

SPSS to determine the sample percentage on your variable of interest along with its sample

size. To do this you should use the SPSS command, FREQUENCIES. Repeat this descriptive

analysis for the other sample, and you will have all the values required (p1, p2, n1, and n2) to

perform the calculations by hand or in a spreadsheet program. (Recall that you can compute

q1 and q2, based on the “p + q = 100%” relationship.)



SPSS does not perform

tests of the significance

of the difference between

the percentages of two

groups, but you can use

SPSS to generate the

relevant information

and perform a hand

calculation.



360



Chapter 13 • ImplementIng BasIC DIfferenCes tests



Marketing research insight 13.2



Global Application



Impulse purchases Differ in French Versus Swedish Consumers

It is well known that consumers buy on impulse, and supermarket

purchases are commonly cited as rife with impulse purchasing. A

recent study7 addressed the question “Is impulse purchasing universal?” meaning does it exist in consumers regardless of their

nationalities. If it is universal, then marketers can use similar impulse purchase strategies, such as end-of-aisle and checkout displays, to stimulate it in supermarkets. The study compared French

and Swedish supermarket shoppers and found the following

percentages of impulse purchases across 15 product categories.

In the figure, bars that do not have percent labels pertain

to those products where no statistically significant (95% level of



confidence) differences were found. It is interesting to note that

while impulse buying is universal for these two nationalities, it

differs by product category. Specifically, French grocery shoppers

are more prone to impulse purchases of stable food items, such as

crackers and biscuits, fruits, and cheese, than are Swedish grocery

shoppers, while Swedish shoppers are more prone to impulse

purchases of snacks such as candy and peanuts and potato chips

as well as soft drinks, than are French shoppers. These differences

findings reveal that marketers who seek to stimulate impulse

shopping in supermarkets must vary their strategies according to

the customs in each country in which they are operating.



Impulse Purchases of French and Swedish Grocery Shoppers



Differences betWeen Means With tWO grOuPs

(inDePenDent saMPLes)

The procedure for testing significance of difference between two means, from two different

groups (either two different samples or two different groups in the same sample) is identical

to the procedure used in testing two percentages. However, the equations differ because a

scale variable is involved.