Tải bản đầy đủ
2 THE RELIABILITY TEST: THE CRONBACH’S ALPHA TEST

2 THE RELIABILITY TEST: THE CRONBACH’S ALPHA TEST

Tải bản đầy đủ

specific item with total of the other items in the scale ( Corrected Item-Total correlation)
is moderately high or higher 0.3 (Burnstein & Nunnally, as cited in Nguyen, 2011), the
item is probably at least moderately correlated with the most of the other item and will
make a good component of this summated rating rate. On the other hand, if the item of
total correlation is negative or too low (less than 0.3), it is required to check again the
words used in questionnaire, take a look on the meaning between each item and modify
the item if it is necessary for conceptual fit.
The results of Cronbach’s alpha test for each construct were summarized in below table:
Table 4.2
The results of Cronbach’ alpha

attitude1
attitude2
attitude3

Scale
Mean if
Corrected Item Item
Scale Variance Total
Cronbach's Alpha
Deleted
if Item Deleted Correlation
Item Deleted
3.9205
3.799
0.604
0.727
3.6875
3.393
0.639
0.685
3.6193
3.094
0.629
0.702

risk1
3.7955
1.295
0.549
risk2
4.2045
1.535
0.549
Perceived risk
Social1
8.0284
4.256
0.67
Social2
8.5568
3.642
0.77
Social3
8.2898
3.738
0.727
Social cost
Purchase1 2.3125
1.508
0.817
Purchase2 2.3125
1.416
0.817
Purchase intention toward counterfeit product

if

Cronbach's Alpha

0.783

Cronbach's Alpha
0.838
0.743
0.786
Cronbach's Alpha

0.707

Cronbach's Alpha

0.851

0.899

29

The result performed that 4 scales had the result of Cronbach’s alpha above 0.6, the
highest was 0.899 (purchase intention toward counterfeit products) and the lowest was
0.707 (perceived risk). Moreover, the corrected item-total correlation of each item is
above 0.3. This indicates that all scales fit the requirement for reliability. As a result,
these measures were used in establishing the main survey to test the study hypotheses.
4.3 EXPLORATORY FACTOR ANALYSIS (EFA)
After analyzing the Cronbach’s alpha, the author evaluated the measurement
scales by conducting exploratory factor analysis. The purpose of EFA is to define which
set of items go together as a group or are answered similarly by respondents (Leech et al.,
2005). In this study, EFA was run through the Principal Axis Factoring with Varimax
rotation method. As the conceptual model that there are four factors: Perceived risk,
social cost, consumers’ attitude toward purchasing counterfeited products and purchase
intention toward counterfeit products. The author examined if the items belonging to one
concept actually are in the same group.
Based on the test of assumption, the KMO was 0.771 presenting sufficient items
for each factor. KMO test indicates one whether or not enough items are predicted by
each factor. The Bartlett was significant (0.000 less than 5%) means that the variable are
correlated highly enough to provide a reasonable basis for factor analysis. (See Table
4.3).
By doing EFA (Principal Axis Factoring with Varimax rolation method), the result
showed that four factors were extracted from 10 items measuring: perceived risk, social
cost, consumers’ attitude toward purchasing counterfeit products and purchase intention
toward counterfeit products. Moreover, the cumulative of the first four factors occupied

30

for 78.3 percent of variance. This indicated that nearly eighty percent of variance could
be explained by four initial items.
The Rotated Factor Matrix showed the items and factor loading for rotated factors
with loading higher than 0.5 are significant as requirement. The items clustered into four
groups that they belong to.
Table 4.5
Rotated Component Matrix
Component
Factor
1
Social cost

Social2

0.839

Social3

0.78

Social1

0.683

2

3

Attitude
toward
counterfeit
products

Attitude2

0.709

Attitude1

0.707

Attitude3

0.694

Purchase
intention
toward
counterfeits

Purchase2

0.844

Purchase1

0.802

4

Risk2

0.701

Risk1

0.677

Perceived risk

4.4 MULTIPLE REGRESSION ANALYSIS

31

After testing the Cronbach’s Alpha Analysis and EFA, the author conducted the
multiple regression analysis in order to define the relationship between four factors
mentioned above. According to Hair et al. (2010), multiple regression analysis helps the
author to predict the level of impact of independent variable on dependent variable.
Therefore, basing on the proposed conceptual model, the author ran two times of multiple
regression analysis. The first time was to justify the impact of social cost and perceived
risk on the consumers’ attitude toward purchasing counterfeit products. The second time
was to evaluate the influence of social cost, perceived risk and the consumers’ attitude
toward purchasing counterfeit products on purchase intention toward counterfeits.
In order to make sure the multiple regression performs exactly analysis result, it is
necessary to test these following main assumptions:
-

Assumption 1: The residuals were independent.

-

Assumption 2: There was the linear relationship between independent variables
and dependent variable.

-

Assumption 3: The residual or error was distributed normally.

-

Assumption 4: No multicollinearity. This assumption is important to test when
and/ or inaccurate results. Multicollinearity happens when there are high
intercorrelations among some set of the predictor of the predictor varibles
(Leech et al., 2005). Multicollinearity could be checked from a correlation
matrix and coefficents results.

32

4.4.1 Multiple regression analysis to define the impact of perceived risk, social cost
on the consumers’ attitude toward counterfeited products.
Results of testing assumptions
Assumption 1:
In order to check this assumption, the Durbin-Watson value was examined. If DurbinWaston was around 2, the residuals could be concluded to be independent. As table 4.8, the value
of Durbin-Watson was 1.984. Therefore, it was qualified with assumption 1.

Assumption 2:
This assumption could be checked by using Curve Estimation Graphs and Scatter
Graphs to define the relationship between consumers’ attitude toward purchasing
counterfeit products and each predictor. As the plot shown in Appendix C, the
relationship between consumers’ attitude toward purchasing counterfeit products and
each predictor was linear regression. Therefore, this assumption was satisfied.
Assumption 3:
The normal distribution of dependent variable was test by histogram and scatter
plot (as Appendix C). The result of this test was support for the normal distribution.
Assumption 4:
Mulicollinearity happens when there are the high inter-correlations among some
composite of the independent variables. According to the correlation matrix below,
correlations among predictors were low and the Pearson test indicated the value lower
than 0.8. This means that there is a low possibility with multicollinearity.

33

Table 4.6
Correlations Matrix of Social, Risk and Attitude
attitude
Attitude

Risk

social

1

Risk

-.147

Social

-.221**

1
.473**

1

**. Correlation is significant at the 0.01 level (2-tailed).

In the other hand, assumption about multicollinearity is also tested by evaluating
two collinearity diagnostic factors (tolerance and the Variance Inflation Factor – VIF).
The Variance Inflation Factor (VIF) performs the influence of collinearity among the
variables in a regression model. VIF is always greater than or equal to 1. When tolerance
value is less than 0.1 and VIF >2 indicates that there is high possibility of
multicollinearity. In this research, the result showed that all tolerances were higher than
0.1 and VIF less than 2. Thus, there was not multicollinearity. This assumption is
satisfied.
In brief, collected data was satisfied with 3 main assumptions to run multiple
regression test.
After checking all assumptions, the result of running Multiple regression was reported to
determine how well the model fit.

34

Table 4.7
Model Summary
Mode
l
R
1

R Square
Square

.227a

.051

Std. Error of the
Estimate

.040

Durbin-Watson

2.58413

1.984

a. Predictors: (Constant), social, risk
b. Dependent Variable: attitude

According to the Model Summary table, the multiple correlation coefficient (R)
was 0.227, R Square was equal 0.051 and adjusted R Square was 0.04, showing that 5,1%
of the variance in consumers’ attitude toward purchasing counterfeit products could be
predicted from two independent variables.
Table 4.8
ANOVA
Sum of
Squares

Model
1

Regression

df

Mean
Square

F
4.678

62.479

2

31.239

Residual

1155.248

173

6.678

Total

1217.727

175

Sig.
.011a

a. Predictors: (Constant), social, risk
b. Dependent Variable: attitude

The value of F was 4.678 and sig<0.05 indicates that the combination of these variables
significantly predicts the dependent variable.

35

Table 4.9
Regression result of “attitude” model

Model

1

Unstandardized
Coefficients

Standardized
Coefficients

B

Beta

Std. Error

Collinearity
Statistics
T

Sig.

8.789

.000

Tolerance VIF

(Constant) 8.415

.957

Risk

-.069

.106

-.054

-.647

.518

.776

1.289

Social

-.181

.078

-.196

-2.327

.021

.776

1.289

a. Dependent Variable: attitude

H2: Perceive risk is negatively related to consumers’ attitude toward purchasing
counterfeit products.
According to Coefficients matrix, the factor of perceived risk had sig value around
0.518 which was higher 0.05 as requirement. This meant consumers who perceive more
or less risk in purchasing counterfeit products will not impact on their attitude toward
purchasing these products. Therefore, the hypothesis 1 was not supported. The difference
between the finding of this study and previous research was discussed on following
chapter.
H4: Social cost of counterfeits is negatively related to consumers’ attitude toward
In order to evaluate the impact of social cost, the result showed that the fact of
social cost was negatively related to consumers’ attitude toward purchasing pirated goods
because its Standardized Coefficient was negative (-0.196) and sig, 0.021<0.05. This

36

implied that this factor which investigated in this research was a meaningful factor to
consumers’ attitude toward purchasing counterfeit products and that if consumers are
unfavorable for purchasing these goods. So, the hypothesis 3 was well confirmed.
4.4.2 Multiple regression analysis to examine the influence of perceived risk, social
intention toward pirated goods.
The result of all assumptions
Assumption 1
As Table 4.13, the value of Durbin-Watson was 1.793 nearly 2. Therefore, it was
qualified with assumption 1.
Assumption 2
As the plot shown in Appendix D, the relationship between purchase intention toward
counterfeit products and each predictor which includes factors of perceived risk, social
cost and consumers’ attitude toward purchasing counterfeit products was linear
regression. Therefore, this assumption was qualified.
Assumption 3:
The residuals or errors had normally distributed (Mean = 3.04E-16 nearly 0 and
Standard deviation Std. Dev = 0.991 nearly 1. Therefore, the collected data was met the
requirement of this assumption.
Assumption 4

37

It is obligated to test correlation between variables by using Pearson correlations
(see table 4.11). The results indicated that explanatory variables are not corrected with
each other.
Table 4.10
Correlations Matrix of Purchase, Social, Risk and Attitude

Discription
Attitude
Risk
Social
Purchase

attitude

risk

Social Purchase

1
-.147

1

-.221**

.473**

1

.552**

-.199**

-.293**

1

**. Correlation is significant at the 0.01 level (2-tailed).

Similarly, multicollinearity also is checked by evaluating Tolerance and VIF from
Cofficients matrix. As table 4.12, all the VIF values in our analysis were less than 2
(from 1.054 to 1.329) and tolerance values were higher than 0.1. Therefore,
multicollinearity really did not happen in the collected data.
Next, the author went to analyze deeply the multiple regression result to answer
for proposed hypotheses after test all qualified assumptions.

38

Table 4.11
Model Summary

Model

R
.581a

1

R Square

Square

.338

Std. Error of
the Estimate

.326

DurbinWatson

1.89156

1.793

a. Predictors: (Constant), attitude, risk, social
b. Dependent Variable: purchase

The adjusted R square was 0.338 indicated that 33.8% of the variance of purchase
intention toward counterfeit products could be predicted from the independent variable.
Therefore, the model of this study was quite fit.

Table 4.12
ANOVA
Sum of
Squares

df

Mean
Square

F

Regression 313.831

3

104.610

29.237

Residual

615.419

172

3.578

Total

929.250

175

Model
1

Sig.
.000a

a. Predictors: (Constant), attitude, risk, social
b. Dependent Variable: purchase

The value of F was 29.237 and sig<0.05 indicates that the combination of perceived risk,
social cost and consumers’ attitude toward purchasing counterfeit products meaningfully
predicts the dependent variable.

39