Download as pdf or txt
Download as pdf or txt
You are on page 1of 15

Discriminant Analysis

Take Home Exercise


Q.1:

a.)

Wilks' Lambda
Test of Wilks'
Function(s) Lambda Chi-square df Sig.
1 .533 16.690 3 <.001

Here, the p-value is < 0.001, which is less than α (0.05), so the discriminant model is significant. So, it
means the discriminant model significantly explains the variation in Loyalty to the Brand (dependent
variable).
The interpretation is that Attitude Toward the Brand, Attitude Toward the Product Category, and
Attitude Toward Shopping (Independent Variables) in the discriminant model significantly explain the
variation in Loyalty to the Brand (dependent variable).
Here, Wilk’s Lambda is 0.533, which is moderate, thus indicating that the group means of the
discriminant model may not be different.

Eigenvalues
Eigenvalu % of Cumulative Canonical
Function e Variance % Correlation
1 .877 a 100.0 100.0 .684
a. First 1 canonical discriminant functions were used in the
analysis.

Eigen Value = 0.877, which is not so high so it is not that better.

Canonical Correlation = 0.684, which indicates a moderate correlation between discriminant score
and corresponding group membership (Loyalty to the Brand – Loyal/ Not Loyal)
Square of canonical correlation = 0.4678
So, 46.78% of the variance of the dependent variable (Loyalty to the Brand) is explained by
independent variables (Attitude Toward the Brand, Attitude Toward the Product Category, and
Attitude Toward Shopping) i.e. model.
Tests of Equality of Group Means
Wilks'
Lambda F df1 df2 Sig.
Attitude Toward the .558 22.192 1 28 <.001
Brand
Attitude Toward the .985 .438 1 28 .514
Product Category
Attitude Toward the .842 5.237 1 28 .030
Shopping

Here, from the p-value, the Attitude Toward the Brand and Attitude Toward the Shopping is
significant [less than α (0.05)], and the Attitude Toward the Product Category is not significant [more
than α (0.05)].

And, from Wilk’s Lambda, Attitude Toward the Brand has a lower value indicating that for this, group
the means seem to be different and Attitude Toward the Product Category and Attitude Toward the
Shopping has a higher value indicating that for this, group the means may not be different.

Group Statistics
Valid N (listwise)
Std. Unweighte
Loyalty to the Brand Mean Deviation d Weighted
Loyal Attitude Toward the 5.67 1.234 15 15.000
Brand
Attitude Toward the 4.20 1.612 15 15.000
Product Category
Attitude Toward 4.00 1.000 15 15.000
Shopping
Not Loyal Attitude Toward the 3.53 1.246 15 15.000
Brand
Attitude Toward the 3.80 1.699 15 15.000
Product Category
Attitude Toward 3.07 1.223 15 15.000
Shopping
Total Attitude Toward the 4.60 1.632 30 30.000
Brand
Attitude Toward the 4.00 1.640 30 30.000
Product Category
Attitude Toward 3.53 1.196 30 30.000
Shopping
From this table, it could be analysed that the means of Attitude Toward the Brand is significantly
higher for Loyal group than Not Loyal group. And, the means of Attitude Toward the Product
Category and Attitude Toward the Shopping is not significantly different for Loyal group than Not
Loyal group.

Canonical Discriminant
Function Coefficients
Function
1
Attitude Toward the .708
Brand
Attitude Toward the .078
Product Category
Attitude Toward .286
Shopping
(Constant) -4.583
Unstandardized coefficients

From here, the discriminant score equation could be written as


Discriminant = -4.583 + 0.708*Attitude Toward the Brand + 0.078*Attitude Toward the Product
Category + 0.286*Attitude Toward Shopping
• β1 = 0.708. This means, everything else remaining constant, when there is one unit
increase in Attitude Toward the Brand, the Discriminant score increases by 0.708
unit.
• β2 = 0.078. This means, everything else remaining constant, when there is one unit
increase in Attitude Toward the Product Category, the Discriminant score increases
by 0.078 unit.
• Β3 = 0.286. This means, everything else remaining constant, when there is one unit
increase in Attitude Toward Shopping, the Discriminant score increases by 0.286 unit.

Standardized Canonical
Discriminant Function
Coefficients
Function
1
Attitude Toward the .879
Brand
Attitude Toward the .130
Product Category
Attitude Toward .320
Shopping

From the table, Attitude Toward the Brand has the highest relative contribution among the variables
in discriminating between the two groups (Loyalty to the Brand).

Pooled Within-Groups Matrices


Attitude
Attitude Toward the Attitude
Toward the Product Toward
Brand Category Shopping
Correlation Attitude Toward the 1.000 .080 .193
Brand
Attitude Toward the .080 1.000 -.209
Product Category
Attitude Toward .193 -.209 1.000
Shopping
From this table, we can say that the multi-collinearity (correlation between independent variables) is
less in the model.

Structure Matrix
Function
1
Attitude Toward the .950
Brand
Attitude Toward .462
Shopping
Attitude Toward the .133
Product Category
Pooled within-groups correlations
between discriminating variables and
standardized canonical discriminant
functions
Variables ordered by absolute size
of correlation within function.

From the table, we can see that Attitude Toward the Brand has a high correlation with discriminant
score, Attitude Toward the Shopping has a moderate correlation with discriminant score, and
Attitude Toward the Product Category has a low correlation with discriminant score.
Functions at Group
Centroids
Loyalty to the Function
Brand 1
Loyal .905
Not Loyal -.905
Unstandardized canonical
discriminant functions
evaluated at group means

This table gives the means of discriminant scores for the two groups (Loyal and Not Loyal). Both are
the same in this case.

Cut-off score = (-0.905+ 0.905)/2 = 0 (Sample size is same for both groups)

Classification Resultsa,c
Predicted Group
Loyalty to the Membership
Brand Loyal Not Loyal Total
Original Count Loyal 12 3 15
Not Loyal 3 12 15
% Loyal 80.0 20.0 100.0
Not Loyal 20.0 80.0 100.0
Cross- Count Loyal 11 4 15
validated b
Not Loyal 3 12 15
% Loyal 73.3 26.7 100.0
Not Loyal 20.0 80.0 100.0
a. 80.0% of original grouped cases correctly classified.
b. Cross validation is done only for those cases in the analysis. In cross validation,
each case is classified by the functions derived from all cases other than that case.
c. 76.7% of cross-validated grouped cases correctly classified.

Hit Ratio = (12+12)/30 = 0.8 or 80%


b.)

In the previous model, Attitude Toward the Product Category is not significant, so the model could be
adjusted to remove the insignificant IV.

Wilks' Lambda
Test of Wilks'
Function(s) Lambda Chi-square df Sig.
1 .537 16.804 2 <.001
Here, the p-value is < 0.001, which is less than α (0.05), so the discriminant model is significant. So, it
means the discriminant model significantly explains the variation in Loyalty to the Brand (dependent
variable).
The interpretation is that Attitude Toward the Brand and Attitude Toward Shopping (Independent
Variables) in the discriminant model significantly explain the variation in Loyalty to the Brand
(dependent variable).
Here, Wilk’s Lambda is 0.537, which is moderate, thus indicating that the group means of the
discriminant model may not be different.

Eigenvalues
Eigenvalu % of Cumulative Canonical
Function e Variance % Correlation
1 .863 a 100.0 100.0 .681
a. First 1 canonical discriminant functions were used in the
analysis.

Eigen Value = 0.863, which is not so high so it is not that better.

Canonical Correlation = 0.681, which indicates a moderate correlation between discriminant score
and corresponding group membership (Loyalty to the Brand – Loyal/ Not Loyal)
Square of canonical correlation = 0.4637
So, 46.37% of the variance of the dependent variable (Loyalty to the Brand) is explained by
independent variables (Attitude Toward the Brand and Attitude Toward Shopping) i.e. model.

Tests of Equality of Group Means


Wilks'
Lambda F df1 df2 Sig.
Attitude Toward the .558 22.192 1 28 <.001
Brand
Attitude Toward .842 5.237 1 28 .030
Shopping

Here, from the p-value, the Attitude Toward the Brand and Attitude Toward the Shopping is
significant [less than α (0.05)].

And, from Wilk’s Lambda, Attitude Toward the Brand has a lower value indicating that for this, group
the means seem to be different and Attitude Toward Shopping has a higher value indicating that for
this, group the means may not be different.

Group Statistics
Valid N (listwise)
Std. Unweighte
Loyalty to the Brand Mean Deviation d Weighted
Loyal Attitude Toward the 5.67 1.234 15 15.000
Brand
Attitude Toward 4.00 1.000 15 15.000
Shopping
Not Loyal Attitude Toward the 3.53 1.246 15 15.000
Brand
Attitude Toward 3.07 1.223 15 15.000
Shopping
Total Attitude Toward the 4.60 1.632 30 30.000
Brand
Attitude Toward 3.53 1.196 30 30.000
Shopping

From this table, it could be analysed that the means of Attitude Toward the Brand is significantly
higher for Loyal group than Not Loyal group. And, the means of Attitude Toward the Shopping is not
significantly different for Loyal group than Not Loyal group.

Canonical Discriminant
Function Coefficients
Function
1
Attitude Toward the .727
Brand
Attitude Toward .261
Shopping
(Constant) -4.268
Unstandardized coefficients

From here, the discriminant score equation could be written as


Discriminant = -4.268 + 0.727*Attitude Toward the Brand + 0.261*Attitude Toward Shopping
• β1 = 0.727. This means, everything else remaining constant, when there is one unit
increase in Attitude Toward the Brand, the Discriminant score increases by 0.727
unit.
• β2 = 0.261. This means, everything else remaining constant, when there is one unit
increase in Attitude Toward Shopping, the Discriminant score increases by 0.261 unit.

Standardized Canonical
Discriminant Function
Coefficients
Function
1
Attitude Toward the .902
Brand
Attitude Toward .292
Shopping

From the table, Attitude Toward the Brand has the highest relative contribution among the variables
in discriminating between the two groups (Loyalty to the Brand).

Pooled Within-Groups Matrices


Attitude Attitude
Toward the Toward
Brand Shopping
Correlation Attitude Toward the 1.000 .193
Brand
Attitude Toward .193 1.000
Shopping

From this table, we can say that the multi-collinearity (correlation between independent variables) is
less in the model.
Structure Matrix
Function
1
Attitude Toward the .958
Brand
Attitude Toward .465
Shopping
Pooled within-groups correlations
between discriminating variables
and standardized canonical
discriminant functions
Variables ordered by absolute size
of correlation within function.

From the table, we can see that Attitude Toward the Brand has a high correlation with discriminant
score and Attitude Toward the Shopping has a moderate correlation with discriminant score.

Functions at Group
Centroids
Loyalty to the Function
Brand 1
Loyal .898
Not Loyal -.898
Unstandardized canonical
discriminant functions
evaluated at group means

This table gives the means of discriminant scores for the two groups (Loyal and Not Loyal). Both are
the same in this case.

Cut-off score = (-0.898+ 0.898)/2 = 0 (Sample size is same for both groups)

Classification Resultsa,c
Predicted Group
Loyalty to the Membership
Brand Loyal Not Loyal Total
Original Count Loyal 12 3 15
Not Loyal 3 12 15
% Loyal 80.0 20.0 100.0
Not Loyal 20.0 80.0 100.0
Count Loyal 12 3 15
Cross- Not Loyal 3 12 15
validatedb % Loyal 80.0 20.0 100.0
Not Loyal 20.0 80.0 100.0
a. 80.0% of original grouped cases correctly classified.
b. Cross validation is done only for those cases in the analysis. In cross validation,
each case is classified by the functions derived from all cases other than that case.
c. 80.0% of cross-validated grouped cases correctly classified.

Hit Ratio = (12+12)/30 = 0.8 or 80%

c.)

From the Excel file, the estimation and validation sample have the same Hit rate.

Wilks' Lambda
Test of Wilks'
Function(s) Lambda Chi-square df Sig.
1 .481 12.082 3 .007

Here, the p-value is 0.007, which is less than α (0.05), so the discriminant model is significant. So, it
means the discriminant model significantly explains the variation in Loyalty to the Brand (dependent
variable).
The interpretation is that Attitude Toward the Brand, Attitude Toward the Product Category, and
Attitude Toward Shopping (Independent Variables) in the discriminant model significantly explain the
variation in Loyalty to the Brand (dependent variable).
Here, Wilk’s Lambda is 0.481, which is low, thus indicating that the group means of the discriminant
model seem to be different.

Eigenvalues
Eigenvalu % of Cumulative Canonical
Function e Variance % Correlation
1 1.080 a 100.0 100.0 .721
a. First 1 canonical discriminant functions were used in the
analysis.
Eigen Value = 1.080, which is not so high so it is not that better.

Canonical Correlation = 0.721, which indicates a high correlation between discriminant score and
corresponding group membership (Loyalty to the Brand – Loyal/ Not Loyal)
Square of canonical correlation = 0.5198
So, 51.98% of the variance of the dependent variable (Loyalty to the Brand) is explained by
independent variables (Attitude Toward the Brand, Attitude Toward the Product Category, and
Attitude Toward Shopping) i.e. model.

Tests of Equality of Group Means


Wilks'
Lambda F df1 df2 Sig.
Attitude Toward the .624 10.866 1 18 .004
Brand
Attitude Toward the .915 1.664 1 18 .213
Product Category
Attitude Toward .757 5.786 1 18 .027
Shopping
Here, from the p-value, the Attitude Toward the Brand and Attitude Toward the Shopping is
significant [less than α (0.05)], and the Attitude Toward the Product Category is not significant [more
than α (0.05)].

And, from Wilk’s Lambda, Attitude Toward the Brand has a moderate value indicating that for this,
group the means seem to be different and Attitude Toward the Product Category and Attitude
Toward the Shopping has a higher value indicating that for this, group the means may not be
different.

Group Statistics
Valid N (listwise)
Std. Unweighte
Loyalty to the Brand Mean Deviation d Weighted
Loyal Attitude Toward the 5.67 1.234 15 15.000
Brand
Attitude Toward the 4.20 1.612 15 15.000
Product Category
Attitude Toward 4.00 1.000 15 15.000
Shopping
Not Loyal Attitude Toward the 3.60 1.140 5 5.000
Brand
Attitude Toward the 3.00 2.345 5 5.000
Product Category
Attitude Toward 2.80 .837 5 5.000
Shopping
Total Attitude Toward the 5.15 1.496 20 20.000
Brand
Attitude Toward the 3.90 1.832 20 20.000
Product Category
Attitude Toward 3.70 1.081 20 20.000
Shopping

From this table, it could be analysed that the means of Attitude Toward the Brand is significantly
higher for Loyal group than Not Loyal group. And, the means of Attitude Toward the Product
Category and Attitude Toward the Shopping is not significantly different for Loyal group than Not
Loyal group.

Canonical Discriminant
Function Coefficients
Function
1
Attitude Toward the .579
Brand
Attitude Toward the .209
Product Category
Attitude Toward .691
Shopping
(Constant) -6.354
Unstandardized coefficients

From here, the discriminant score equation could be written as


Discriminant = -6.354 + 0.579*Attitude Toward the Brand + 0.209*Attitude Toward the Product
Category + 0.691*Attitude Toward Shopping
• β1 = 0.579. This means, everything else remaining constant, when there is one unit
increase in Attitude Toward the Brand, the Discriminant score increases by 0.579
unit.
• β2 = 0.209. This means, everything else remaining constant, when there is one unit
increase in Attitude Toward the Product Category, the Discriminant score increases
by 0.209unit.
• Β3 = 0.691. This means, everything else remaining constant, when there is one unit
increase in Attitude Toward Shopping, the Discriminant score increases by 0.691 unit.

Standardized Canonical
Discriminant Function
Coefficients
Function
1
Attitude Toward the .703
Brand
Attitude Toward the .377
Product Category
Attitude Toward .667
Shopping

From the table, Attitude Toward the Brand has the highest relative contribution among the variables
in discriminating between the two groups (Loyalty to the Brand).

Pooled Within-Groups Matrices


Attitude
Attitude Toward the Attitude
Toward the Product Toward
Brand Category Shopping
Correlation Attitude Toward the 1.000 .152 -.019
Brand
Attitude Toward the .152 1.000 -.287
Product Category
Attitude Toward -.019 -.287 1.000
Shopping

From this table, we can say that the multi-collinearity (correlation between independent variables) is
less in the model.
Structure Matrix
Function
1
Attitude Toward the .748
Brand
Attitude Toward .546
Shopping
Attitude Toward the .293
Product Category
Pooled within-groups correlations
between discriminating variables and
standardized canonical discriminant
functions
Variables ordered by absolute size
of correlation within function.

From the table, we can see that Attitude Toward the Brand has a high correlation with discriminant
score, Attitude Toward the Shopping has a moderate correlation with discriminant score, and
Attitude Toward the Product Category has a low correlation with discriminant score.

Functions at Group
Centroids
Loyalty to the Function
Brand 1
Loyal .569
Not Loyal -1.707
Unstandardized canonical
discriminant functions
evaluated at group means

This table gives the means of discriminant scores for the two groups (Loyal and Not Loyal). Both are
the same in this case.

Cut-off score = (5* 0.5688+15* -1.7078)/ (15+5) = -1.13865 (Sample size is different for both groups)
Classification Resultsa,c
Predicted Group
Loyalty to the Membership
Brand Loyal Not Loyal Total
Original Count Loyal 14 1 15
Not Loyal 1 4 5
% Loyal 93.3 6.7 100.0
Not Loyal 20.0 80.0 100.0
Cross- Count Loyal 12 3 15
validated b
Not Loyal 1 4 5
% Loyal 80.0 20.0 100.0
Not Loyal 20.0 80.0 100.0
a. 90.0% of original grouped cases correctly classified.
b. Cross validation is done only for those cases in the analysis. In cross validation,
each case is classified by the functions derived from all cases other than that case.
c. 80.0% of cross-validated grouped cases correctly classified.

Hit Ratio = (14+4)/20 = 0.9 or 90%

You might also like