QTA

Discriminant Analysis

Discriminant Analysis
Analysis of Dependence
We now focus our discussion on the multivariate techniques that deal with
analysis of dependence.
The purpose of these techniques is to predict a variable from a set of
independent variables. The dependence techniques we cover in this course
include multiple regression, discriminant analysis, and cluster analysis.

Discriminant Analysis

Discriminant analysis is a multivariate technique used for predicting group

membership on the basis of two or more independent variables.

There are many situations where the marketing researchers purpose is to

classify objects or groups by a set of independent variables.

The dependent variable in discriminant analysis is non-metric or categorical.

Discriminant analysis: A technique for analyzing marketing research

data when the criterion or dependent variable is categorical and the
predictor or independent variables are intervals.
In marketing, consumers are often categorized on the basis of heavy versus
light users of a product, or viewers versus non-viewers of a media vehicle such
as a television commercial.

The independent variables in discriminant analysis are metric and often

include characteristics such as demographics and psychographics.

Lets begin our discussion of discriminant analysis with an intuitive example.

Our fast-food company wants to see whether a lifestyle variable such as
eating a nutritious meal X1 and a demographic variable such as household
income X2 are useful in distinguishing households visiting their restaurant
from those visiting other fast-food restaurants.

Marketing researchers have gathered data on X1 and X2 for two random

samples of households.

Discriminant analysis procedures would plot these data on a two-dimensional

graph, as shown in figure-1.

The scatter plot in figure-1 yields two groups, one containing primarily Back
Yard Burgers customers and the other containing primarily households that
patronize other fast-food restaurants.

From this example, it appears that X1 and X2 are critical discriminators of fastfood restaurant patronage. Although the two areas overlap, the extent of the
overlap does not seem to be substantial.

This minimal overlap between groups, as in figure-1, is an important

requirement for a successful discriminant analysis.

Let us now turn to the fundamental statistics of discriminant analysis.

Remember, the prediction of a categorical variable is the purpose of
discriminant analysis.

QTA
Discriminant Analysis
From a statistical perspective, this involves studying the direction of group
differences based on finding a linear combination of independent variables
the discriminant functionthat shows large differences in group means.

Thus, discriminant analysis is a statistical tool for determining linear

combinations of those independent variables.

Discriminant function: The linear combination of independent variables

developed by discriminant analysis which will best discriminate between
the categories of the dependent variable.

Figure-I: Discriminant Analysis Scatter Plot of Lifestyle and Income Data for Fast-Food
Restaurant Patronage

A linear function can be developed with our fast-food example.

We will use a two group discriminant analysis example in which the dependent
variable, Y , is measured on a nominal scale (i.e., patrons of Back Yard Burgers
versus other fast-food restaurants).

Again, the marketing manager believes it is possible to predict whether a

customer will patronize a fast-food restaurant on the basis of lifestyle X1 and
income X2 .

Now the researcher must find a linear function of the independent variables
that shows large differences in group means.

The discriminant score, or the Z score, is the basis for predicting to which
group the particular individual belongs and is determined by a linear function.

This Z score will be derived for each individual by means of the following
equation:
Zi b1 X1i b2 X2i bn Xni

Where; Zi = ith individuals discriminant score, bn = Discriminant coefficient for

the nth variable

Xni = Individuals value on the nth independent variable.

QTA

Discriminant Analysis

Discriminant score: In discriminant analysis,

respondent on the discriminant function.

the

score

of

each

Discriminant weights bn , or discriminant function coefficients, are estimates

of the discriminatory power of a particular independent variable.

These coefficients are computed by means of the discriminant analysis

computer program. The size of the coefficients associated with a particular
independent variable is determined by the variance structure of the variables
in the equation.

Independent variables with large discriminatory power will have large weights,
and those with little discriminatory power will have small weights.
Discriminant function coefficients: The multipliers of variables in the
discriminant function when the variables are in the original units of
measurement.

Returning to our fast-food example, suppose the marketing manager finds the
standardized weights or coefficients in the equation to be:

Zi b1 X1 b2 X2
0.32X1 0.37X2

These results show that income X2 is the more important variable in

discriminating between those patronizing Back Yard Burgers and those who
patronize other fast-food restaurants.
The lifestyle variable X1 with a coefficient of .32 also represents a variable
with good discriminatory power.

Another important goal of discriminant analysis is classification of objects or

individuals into groups. In our example, the goal was to correctly classify
consumers into BackYard Burger user and nonuser groups.

To determine whether the estimated discriminant function is a good predictor,

a classification (prediction) matrix is used.
Predicted Group
User
Nonuser
User
93
17
Actual
(84.5% (15.5%)
Group
)
Nonus 27
95
er
(22.1% (77.9%)
)
Overall prediction accuracy = 81% (93 + 95
= 188/232 )

Total
110

122

N = 232

Table-I: Classification Matrix for Users and Nonusers

3

QTA

Discriminant Analysis

The classification matrix in Table-I shows that the discriminant function

correctly classified 84.5 percent of the actual users as users and incorrectly
classified 15.5 percent of the actual users as nonusers.

Regarding nonusers, the discriminant function correctly classified 77.9 percent

as nonusers and incorrectly classified 22.1 percent as users.

Our classification matrix in Table-I shows that the number of correctly

classified consumers (93 users and 95 nonusers) out of a total of 232 equals
81 percent correctly classified. This resulting percentage is much higher than
would be expected by chance.

Statistical tests can be performed to determine whether the percentage of

correct classification is statistically significant.
Classification (or prediction) matrix: The classification matrix in
discriminant analysis that contains the number of correctly classified
and misclassified cases.

Discriminant Analysis Applications in Marketing

Research

While our example illustrated how discriminant analysis helped classify users
and nonusers of the restaurant based on independent variables, other
applications include the following:

Product Research: Discriminant analysis can help to distinguish between

Image Research: Discriminant analysis can discriminate between customers

Advertising Research: Discriminant analysis can assist in distinguishing how

Direct

heavy, medium, and light users of a product in terms of their consumption

habits and lifestyles.
who exhibit favorable perceptions of a store or company and those who do
not.
market segments differ in media consumption habits.

Marketing:

Discriminant analysis can help in distinguishing

characteristics of consumers who respond to direct marketing solicitations and
those who dont.

SPSS ApplicationDiscriminant Analysis

The usefulness of discriminant analysis can be demonstrated with our Deli

Depot database.

Remember that with discriminant analysis the single dependent variable is a

non-metric variable and the multiple independent variables are measured
metrically.

In the database, variables X7Gender and X10Usage Level are non-metric

variables, as is X11Market Area. Variables X7 and X10 are two-group
variables and X11 is a three-group variable.

We could use discriminant analysis to see if there are differences between

perceptions of Deli Depot by male and female customers or between heavy

QTA
Discriminant Analysis
users and light users, or we could see if the perceptions differ depending on
how far customers drove to eat at Deli Depot.

Lets look at the possible influence of X7Gender. The task is to determine if

perceptions of Deli Depot, as measured by variables X1X6, are different
between males and females. Another way of stating this is Can the
perceptions of Deli Depot predict whether individuals are male or
female?

The SPSS click-through sequence is ANALYZE CLASSIFY DISCRIMINANT,

which leads to a dialog box where you select the variables.

The dependent, non-metric variable is X7 and the independent, metric

variables are X1 to X6, the perceptions variables.

The first thing you do is transfer variable X7 to the Grouping Variable box at
the top, and then click on the Define Range box just below it.

You must tell the program what the minimum and maximum numbers are for
the grouping variable.

In this case the minimum is 0 = female and the maximum is 1 = male, so just
put these numbers in and click on Continue.

Next you must transfer variables X1X6 into the Independents box. Then
click on the Statistics box at the bottom and check Means, Univariate
ANOVAS, and Continue. The Method default is Enter, and we will use this.

Now click on Classify and Compute from group sizes. We do not know if
the sample sizes are equal, so we must check this option. You should also click
Summary Table and then Continue.

We do not use any options under Save, so click OK to run the program.

Discriminant analysis is a program that gives you a lot of output you will not
use for a simple analysis like this one. We will look at only five tables.

The first important information to consider is in the Wilks Lambda table.

The Wilks Lambda is a statistic that assesses whether the discriminant
analysis is statistically significant. If this statistic is significant, and it is in our
case (.000), then we next look at the Classification Results table.

At the bottom we see that the overall ability of our discriminant function to
predict group membership is 90 percent. This is very good because without
the discriminant function we could predict with only 60 percent accuracy (our
sample sizes are males = 20 and females = 30, so if we said all respondents
were female, we would predict with 60 percent accuracy).

Test of
Function(s)
1

Wilks
ChiLambda
Square
.317
51.687
Table-II: WIlks Lambda
Classification
Results

df

Sig.

.000

QTA
Origin
al

Gend
Femal
er
e
Coun Femal 26
t
e
%
Male
1
Femal 86.7
e
Male
5.0
Table-III: Classification

Discriminant Analysis
Male
Total
4

30

19
13.3

20
100.0

95.0
Results

100.0

To find out which of the independent variables help us to predict group

membership we look at the information in two tables.

Results shown in the table labeled Tests of Equality of Group Means show
which perceptions variables differ between males and females on a univariate
basis. Note that variables X1, X2, X3, and X6 are all highly statistically
significant. Variables X4 and X5 are not significant. To consider the variables
from a multivariate perspective, we can either look at the information in the
Standardized Canonical Discriminant Function Coefficients table or in
the Structure Matrix table.

Lets use the information from the Structure Matrix table. First we identify
the numbers in the Function column that are .30 or higher. This cutoff level is
determined in a manner similar to a factor loading.

All variables .30 or higher are considered to be helpful in predicting group

membership. Like the univariate results, the same variablesX1, X3, and X6,
but not X2help us to predict group membership. Variable X2 helps, but it is
not as important a predictor as the other three variables.

X1Friendly
Employees
X2Competitive Price
X3Competent
Employees
X4Food Quality
X5Food Variety
X6Speed of Service

Wilks
Lambda
.679

df1

22.73
1
5
.855
8.140
1
.578
35.10
1
9
1.000
.010
1
1.000
.004
1
.520
44.37
1
7
Table-IV: Tests of Equality of Group Means

X6Speed of Service
X3Competent
Employees
X1Friendly Employees
X2Competitive Price
6

Functio
n
1
.655
.583
.469
.281

df2

Sig.

48

.000

48
48

.006
.000

48
48
48

.920
.947
.000

QTA

Discriminant Analysis
X4Food Quality
.010
X5Food Variety
.007
Table-V: Structure Matrix

As noted earlier, the predictive outcome of the discriminant analysis is shown

in the Classification Results table.

It shows that our discriminant function correctly predicted females 86.7

percent of the time and males 95 percent of the time. Overall, the predictive
accuracy is 90 percent, as shown at the bottom of this table.

To further interpret the discriminant function we look at the group means in

the Group Statistics table for the significant variables. For X1Friendly
Employees we see that females perceived Deli Depots employees to be
significantly more friendly than did males (mean = 4.23 vs. 2.70). Similarly,
females perceived the employees to be more competent (X3) than did males
(mean = 8.77 vs. 6.90). In contrast, females perceived Deli Depots prices (X2)
to be relatively less competitive and speed of service (X6) to be relatively
slower than did males. Thus, perceptions of employees and other operations
variables are significantly different between males and females and are good
predictors of group membership.

The End