Data Analysis
C. Data Analysis
The analysis in this study used Stuctural Equation Model (SEM) statistical methods. In principle, the structural model aimed at testing the causal relationships of variables. In this study, data were analyzed using Analysis of Moment Structure software (AMOS) version 18. There were several assumptions that must be met before performing the test with a structural equation model approach. SEM assumptions included the assumption of sample adequacy, normality, and outliers. The following are the explanations of each assumption.
1. The Assumption of Sampling Adequacy
The samples of this study were 200 respondents. All of questionnaires were possible to be analyzed in this study. The number of respondents met the Maximum Likelihood Estimation procedures for sampling; it is between 100-200 samples (Hair et. al., 1998).
2. The Assumption of Normality
This test aims to determine normality of data distribution pattern to follow or close to normal distribution. Normality is a data distribution in single matrix of variables that produce a normal distribution (Hair et al., 1998). Table IV.4 showed the result of normality test.
Univariate normality was examined by checking the critical ratio of skewness. The data distribution is considered to follow the univariate normality if the absolute value of critical ratio is <2.58. Based on Table IV.4, the data spread of the indicators has followed normal distribution because all of critical ratio absolute value is under 2.58. Therefore, the conclusion is the data distribution has followed univariate normality.
However, univariate normality does not give assurance that the data is normal in multivariate conditions (George, 2001). Examining the kurtosis value was needed to ensure multivariate normality. If critical ratio of kurtosis is <7, the data distribution will follow multivariate normality. Based on Table
VI.4, the critical ratio value was lower than 7, thus, it indicated that the data has followed multivariate normality.
Assessment of Normality
Variable Min Max Skew C.R. Kurtosis C.R.
-,116 -,334 LA6
-,039 -,113 LC6
-,106 -,305 LC5
-,106 -,305 LC4
-,095 -,274 N9
-,152 -,438 N8
-,153 -,442 IM1
-,240 -,694 PC5
1,000 5,000 -,058 -,336
-,104 -,301 PC4
1,000 5,000 -,052 -,298
-,144 -,416 AM4
1,000 4,000 -,030 -,173
-,224 -,648 LS4
1,000 5,000 -,007 -,041
-,079 -,227 LA5
-,077 -,222 LA4
-,113 -,327 LC3
-,053 -,154 LC1
-,039 -,111 IM2
-,231 -,666 IM3
-,227 -,656 IM4
2,000 5,000 -,057 -,330
-,231 -,666 PC3
1,000 5,000 -,045 -,259
-,183 -,527 PC2
1,000 5,000 -,037 -,213
-,149 -,429 PC1
1,000 5,000 -,029 -,168
-,113 -,327 N7
1,000 5,000 -,007 -,043
-,153 -,442 N6
1,000 5,000 -,022 -,128
-,152 -,438 N3
1,000 5,000 -,028 -,163
-,037 -,106 AP1
-,226 -,654 AP2
-,224 -,648 AP3
-,226 -,653 LS3
1,000 5,000 -,014 -,082
-,039 -,113 LS2
1,000 5,000 -,022 -,124
-,077 -,222 LS1
-,013 -,036 LA3
-,190 -,549 LA2
1,000 5,000 -,023 -,132
-,087 -,252 LA1
1,000 5,000 -,074 -,425
-,047 -,137 EM3
3,000 5,000 -,001 -,006
-,059 -,170
Assessment of Normality (continued)
Variable
Min
Max Skew C.R. Kurtosis C.R.
-,059 -,170 EM1
1,000 4,000 -,030 -,173
-,224 -,648 AM2
-,227 -,656 AM1
-,224 -,648 Multivariate
Source: Primary data, made base on Appendix 7, 2012
3. The Assumption of Outliers
Outliers are observations that appear to the extreme values which has unique characteristics that are very different from other observations and appear in the form of extreme value for a single variable or combination of variables (Hair et al., 1998). Generally, the treatment of outliers is by removing them from the data and they are not included in subsequent calculations. If there is no particular reason to exclude outliers, the observations can be included in subsequent analyzes. Outliers can be evaluated by the Mahalanobis distance with the degree of freedom number of variables used in the study at the level of p <0.001. In this case the variable in question is the number of items in the measurement model.
This study was used 41 indicator variables. Thus, if there is the Mahalanob is distance greater than χ2 (41; 0.001) = 74.7449 is the value of multivariate outliers. Summary of the Mahalanobis Distance can be seen in
table IV.5 (The complete list of outliers can be seen in Appendix 7).
Summary of Mahalanobis Distance Squared
Observation number Mahalanobis d-squared p1 p2
--------------------- ------- -------
Source: Primary data, made base on Appendix 7, 2012
Table IV.5 showed that no one of the values was categorized as outliers because the Mahalanobis Distances did not exceed the value of χ2 (41; 0.001) = 74.7449. Thus, the number of samples used in data analysis was 200 samples.
4. Goodness of Fit Analysis
Before hypothesis testing was conducted, the first step was to assess the suitability of goodness-of-fit. The evaluation of the goodness-of-fit model of this study was described in Table IV.6. Measurement of the overall goodness- of-fit model indicated that the model proposed in this study cannot be accepted (Chi-Square 910,925; p = 0,000). Therefore, the researchers consider modifying the model to establish an alternative model that had a better goodness-of-fit.
Goodness of Fit Test Result Criteria
Cut-off Value
Result Absolute Fit Indices
Chi Square
Expected low
910,925 Significance Probability
Incremental Fit Indices
Parsimonious Fit Indices
Normed Chi Square (CMIN/DF)
Source: Primary Data, based on Appendix 7, 2012
5. Modification
Modification aims to obtain an acceptable goodness-of-fit model. Through modification indices value, it can be known several possible modifications to the model that can be carried out. Modification indices on output AMOS Version 18 shows that there are several relationships that can
be correlated to get a better fit of research model. The baseline of modification indices in this research is greater than or equal to 4.0. To obtain an acceptable model, researcher estimates the correlation between the error terms that has the highest value of modification indices (the modification indices list can be seen on Appendix 7). If the path is correlated and the model is analyzed, there will be a decrease in the value of chi square, so the be correlated to get a better fit of research model. The baseline of modification indices in this research is greater than or equal to 4.0. To obtain an acceptable model, researcher estimates the correlation between the error terms that has the highest value of modification indices (the modification indices list can be seen on Appendix 7). If the path is correlated and the model is analyzed, there will be a decrease in the value of chi square, so the
Table IV.7 Goodness of Fit Test Result after Modified
Criteria
Before
After Absolute Fit Indices
Chi Square
806,162 Significance Probability
Incremental Fit Indices
Parsimonious Fit Indices
Normed Chi Square (CMIN/DF)
Source: Primary Data, based on Appendix 7 and 8, 2012
Based on Table IV.7, measurement of the overall goodness-of-fit model shows that the modified model can be accepted (Chi-Square 806,162; p = 0,059). It means the next step of SEM analysis can be processed.
1. Hypothesis Testing
After the goodness-of-fit on structural models were fulfilled the criteria, then the next stage was the analysis of structural relationship among construct (hypothesis testing). Relationships among constructs in the research framework were recognized by noticing the value of regression weights. Table IV.8 shows the regression weights of the tested variables.
Table IV.8 Regression Weights
Regression Weights Estimate S.E. C.R. p
Intrinsic Motivation
Learner Autonomy
0,053 4,420*** 0,000 Intrinsic Motivation
Lecturer Support
0,061 6,555*** 0,000 Academic Performance
Intrinsic Motivation
0,084 3,729*** 0,000 Neutralization
Intrinsic Motivation
0,086 -0,741 0,459 Likelihood of Cheating
Intrinsic Motivation
0,135 -0,820 0,412 Academic Performance
Extrinsic Motivation
0,102 0,655 0,512 Neutralization
Extrinsic Motivation
0,102 -0,649 0,516 Likelihood of Cheating
Extrinsic Motivation
0,161 -0,509 0,611 Academic Performance
Amotivation
0,076 -3,035** 0,002 Neutralization
Amotivation
0,078 -0,551 0,582 Likelihood of Cheating
Amotivation
0,122 -0,802 0,422 Neutralization
Academic Performance
0,075 -2,896** 0,004 Likelihood of Cheating Academic Performance
0,128 0,769 0,442 Neutralization
Prior Cheating
0,070 2,442* 0,015 Likelihood of Cheating
Prior Cheating
0,111 0,173 0,863 Likelihood of Cheating
Source: Primary Data, based on Appendix 8, 2012
*significant at level 5% ; ** significant at level 1% ; ***significant at level 0.1 % *significant at level 5% ; ** significant at level 1% ; ***significant at level 0.1 %
2. Discussions
a. Relationship between neutralization and likelihood of cheating (H1). The test result indicated that there is a significant and positive relationship between neutralization and likelihood of cheating ( β = 0.372, CR = 3.112, SE =0.119, p =0.002). This result explains that the higher the neutralization, the higher the likelihood of cheating. The neutralization, which commonly describes as rationalization, is the main factor that influences likelihood of cheating. The deviant behavior, such as cheating or another crime is creating guilt feeling to the subject. The neutralization provides ways to eliminating the guilt. It also uses as a self-defense when the deviance behaviors are questioned by others. Thus, there is a positive and significant relationship between neutralization and likelihood of cheating.
The research finding supports the hypothesis 1 that describes a positive and significant relationship between neutralization and likelihood of cheating. Theoretically, this research also provides support for the previous research that hypothesized the higher the neutralization, the higher the likelihood of cheating as proposed by Smith et. al. (2009).
Learner Autonomy
Lecturer Support
Prior Cheating
Intrinsic Motivation
Extrinsic Motivation Likelihood of Cheating
Academic Performance
Amotivation
-3,035**
Source: Constructed by the writer, based on Appendix 8, 2012
*significant at level 5% ; ** significant at level 1% ; ***significant at level 0.1 %
The test result indicated that there was a positive and non- significant relationship between prior cheating and likelihood of cheating ( β = 0,019, CR = 0,173, SE = 0,111, p = 0,863). It means that the prior cheating is not a variable that is considered important to influence the likelihood of cheating. This phenomenon could happen because prior cheating did not influence likelihood of cheating directly, but influenced indirectly through academic performance and neutralization. This finding could give understanding for the marketers to give attention towards the non-significant relationship in order to create an effective strategy on cheating prevention by avoiding the non-significant pattern.
This research finding does not support the hypothesis 2 that describes a positive and significant relationship between prior cheating and likelihood of cheating. Theoretically, this research does not provide support for the previous research that hypothesized the higher the prior cheating, the higher the likelihood of cheating as proposed by Smith et. al. (2009).
c. Relationship between prior cheating and neutralization (H3).
The test result indicated that there is a positive and significant relationship between prior cheating and neutralization ( β = 0.171, CR = 2.442, SE = 0.070, p = 0.015). The higher the prior cheating, the higher the neutralization. This finding is indicated that prior cheating had positive The test result indicated that there is a positive and significant relationship between prior cheating and neutralization ( β = 0.171, CR = 2.442, SE = 0.070, p = 0.015). The higher the prior cheating, the higher the neutralization. This finding is indicated that prior cheating had positive
The research finding supports the hypothesis 3 that describes a positive and significance relationship between prior cheating and neutralization. Theoretically, this research also provides support for the previous research that hypothesized the higher the prior cheating, the higher the neutralization as proposed by Smith et. al. (2009).
d. Relationship between academic performance and likelihood of cheating (H4).
The test result indicated that there was a non-significant relationship between academic performance and likelihood of cheating ( β = 0,098, CR = 0,769, SE = 0,128, p = 0,442). It means that the academic performance is not a variable that is considered important to influence the likelihood of cheating. This phenomenon could happen because academic performance did not influence likelihood of cheating directly, but influenced indirectly through neutralization. This finding could give understanding for the marketers to give attention towards the non- significant relationship in order to create an effective strategy on cheating prevention by avoiding the non-significant pattern.
This research finding does not support the hypothesis 4 that describes a negative and significant relationship between academic This research finding does not support the hypothesis 4 that describes a negative and significant relationship between academic
e. Relationship between academic performance and neutralization (H5). The test result indicated that academic performance has a negative and significant influence on neutralization ( β = -0.216, CR = -2.896, SE = 0.075, p = 0.004). The higher the academic performance, the lower the neutralization. Students that have high academic performance do not need to cheat. Because they do not cheat, they do not need to neutralize their behavior. Thus, there is a negative and significance relationship between academic performance and neutralization.
The research finding supports the hypothesis 5 that describes a negative and significance relationship between academic performance and neutralization. Theoretically, this research also provides support for the previous research that hypothesized the higher the academic performance, the lower the neutralization as proposed by Smith et. al. (2009).
f. Relationship between amotivation and likelihood of cheating (H6).
The test result indicated that there was a non-significant relationship between amotivation and likelihood of cheating ( β = -0,098, CR = -0,802, SE = 0,122, p = 0,422). It means that the amotivation is not a The test result indicated that there was a non-significant relationship between amotivation and likelihood of cheating ( β = -0,098, CR = -0,802, SE = 0,122, p = 0,422). It means that the amotivation is not a
This research finding does not support the hypothesis 6 that describes a positive and significant relationship between amotivation and likelihood of cheating. Theoretically, this research does not provide support for the previous research that hypothesized the higher the amotivation, the higher the likelihood of cheating as proposed by Smith et. al. (2009).
g. Relationship between amotivation and neutralization (H7).
The test result indicated that amotivation had a non-significant influence on neutralization ( β = -0,043, CR = -0,551, SE = 0,078, p = 0,582). It means that the amotivation is not a variable that is considered important to influence the neutralization. This phenomenon could happen because amotivation did not influence neutralization directly, but influenced indirectly through academic performance. This finding could give understanding for the marketers to give attention towards the non- The test result indicated that amotivation had a non-significant influence on neutralization ( β = -0,043, CR = -0,551, SE = 0,078, p = 0,582). It means that the amotivation is not a variable that is considered important to influence the neutralization. This phenomenon could happen because amotivation did not influence neutralization directly, but influenced indirectly through academic performance. This finding could give understanding for the marketers to give attention towards the non-
h. Relationship between amotivation and academic performance (H8).
The test result indicated that amotivation had a negative and significant influence on academic performance ( β = -0.232, CR = -3.035, SE = 0.076, p = 0.002). The higher the amotivation, the lower the academic performance. This finding indicated that amotivation is an important variable that influence academic performance. Amotivation refers to the absence of motivation to pursue an activity. If someone is in amotivation state, in this context is toward study, he will get difficulties in performing learning activities. This condition will lead to incompetence or inability to obtain a good academic performance.
The research finding supports the hypothesis 8 that describes a negative and significance relationship between amotivation and academic performance. Theoretically, this research also provides support for the previous research that hypothesized the higher the amotivation, the lower the academic performance as proposed by Smith et. al. (2009).
The test result indicated that there was a non-significant relationship between extrinsic motivation and likelihood of cheating ( β=- 0,082, CR = -0,509, SE = 0,161, p = 0,611). It means that the extrinsic motivation is not a variable that is considered important to influence the likelihood of cheating. This phenomenon could happen because extrinsic motivation did not influence likelihood of cheating. This finding could give understanding for the marketers to give attention towards the non- significant relationship in order to create an effective strategy on cheating prevention by avoiding the non-significant pattern.
This research finding does not support the hypothesis 9 that describes a negative and significant relationship between extrinsic motivation and likelihood of cheating. Theoretically, this research provides support for the previous research that found that the extrinsic motivation does not have significant influence on likelihood of cheating (see Smith et. al., 2009).
j. Relationship between extrinsic motivation and neutralization (H10).
The test result indicated that extrinsic motivation had a non- significant influence on neutralization ( β = -0,066, CR = -0,649, SE = 0,102, p = 0,516). It means that the extrinsic motivation is not a variable that is considered important to influence the neutralization. This phenomenon could happen because extrinsic motivation did not influence The test result indicated that extrinsic motivation had a non- significant influence on neutralization ( β = -0,066, CR = -0,649, SE = 0,102, p = 0,516). It means that the extrinsic motivation is not a variable that is considered important to influence the neutralization. This phenomenon could happen because extrinsic motivation did not influence
The research finding does not support the hypothesis 10 that describes a positive and significant relationship between extrinsic motivation and neutralization. Theoretically, this research provides support for the previous research that found the extrinsic motivation does not have significant effect on the neutralization (see Smith et. al., 2009).
k. Relationship between extrinsic motivation and academic performance (H11). The test result indicated that extrinsic motivation had a non- significant relationship on academic performance ( β = 0.067, CR = 0.655, SE = 0.102, p = 0.512). The change on extrinsic motivation does not influence the academic performance. This research finding indicated that the extrinsic motivation does not included to academic performance predictor. This phenomeno n could happen because students’ academic performance is not influenced by extrinsic motivation. Meanwhile, the academic performance is influenced by amotivation and intrinsic motivation. It is reasonable because the extrinsic motivation is a type of motivation that stands between amotivation and intrinsic motivation if it is approached from level of internalization (see Ryan and Deci, 2000).
on academic performance. This finding could give understanding for the marketers to give attention towards the non-significant relationship in order to create an effective strategy on cheating prevention by avoiding the non-significant pattern.
The research finding does not support the hypothesis 11 that describes a negative and significant relationship between extrinsic motivation and academic performance. Theoretically, this research provides support for the previous research that found the extrinsic motivation does not have significant effect on the academic performance (see Smith et. al., 2009).
l. Relationship between intrinsic motivation and likelihood of cheating (H12). The test result indicated that there was a negative and non- significant relationship between intrinsic motivation and likelihood of cheating ( β = -0,111, CR = -0,820, SE = 0,135, p = 0,412). It means that the intrinsic motivation is not a variable that is considered important to influence the likelihood of cheating. This phenomenon could happen because intrinsic motivation did not influence likelihood of cheating directly, but influenced indirectly through academic performance and neutralization. This finding could give understanding for the marketers to give attention towards the non-significant relationship in order to create an l. Relationship between intrinsic motivation and likelihood of cheating (H12). The test result indicated that there was a negative and non- significant relationship between intrinsic motivation and likelihood of cheating ( β = -0,111, CR = -0,820, SE = 0,135, p = 0,412). It means that the intrinsic motivation is not a variable that is considered important to influence the likelihood of cheating. This phenomenon could happen because intrinsic motivation did not influence likelihood of cheating directly, but influenced indirectly through academic performance and neutralization. This finding could give understanding for the marketers to give attention towards the non-significant relationship in order to create an
m. Relationship between intrinsic motivation and neutralization (H13).
The test result indicated that intrinsic motivation had a negative and non-significant influence on neutralization ( β = -0,063, CR = -0,741, SE = 0,086, p = 0,459). It means that the intrinsic motivation is not a variable that is considered important to influence the neutralization. This phenomenon could happen because intrinsic motivation did not influence neutralization directly, but influenced indirectly through academic performance. This finding could give understanding for the marketers to give attention towards the non-significant relationship in order to create an effective strategy on cheating prevention by avoiding the non-significant pattern.
This research finding does not support the hypothesis 13 that describes a negative and significant relationship between intrinsic motivation and neutralization. Theoretically, this research does not provide This research finding does not support the hypothesis 13 that describes a negative and significant relationship between intrinsic motivation and neutralization. Theoretically, this research does not provide
n. Relationship between intrinsic motivation and academic performance (H14). The test result indicated that there was a positive and significant relationship between intrinsic motivation and academic performance ( β= 0.313, CR = 3.729, SE = 0.084, p = 0.000). The higher the intrinsic motivation, the higher the academic performance. This finding indicated that the intrinsic motivation is an important variable that influence academic performance. Intrinsically motivated students will engage to learning process voluntarily. The learning process represents a challenge to their existing competencies and requires them to use their creative capabilities. The process will have favorable impact on their academic performance.
The research finding supports the hypothesis 14 that describes a positive and significant relationship between intrinsic motivation and academic performance. Theoretically, this research also provides support for the previous research that hypothesized the higher the intrinsic motivation, the higher the academic performance as proposed by Smith et. al. (2009).
The test result indicated that lecturer support had a positive and significant influence on intrinsic motivation ( β = 0.400, CR = 6.555, SE = 0.061, p = 0.000). The higher the lecturer support, the higher the intrinsic motivation. This phenomenon could be happened because there was a causal relationship between lecturer support and intrinsic motivation. Supports from the teacher had result in lower level of stress in study so it created a relaxed condition for the students. Because of the supportive environment, the students became more courage to school. Thus, there was
a positive and significant relationship between lecturer support and intrinsic motivation. The research finding supports the hypothesis 15 that describes a positive and significant relationship between lecturer support and intrinsic motivation. Theoretically, this research also provides support for the previous research that hypothesized the higher the lecturer support, the higher the intrinsic motivation as proposed by Soenens and Vansteenkiste (2005).
p. Relationship between learner autonomy and intrinsic motivation (H16).
The test results indicated that learner autonomy had a positive and significant influence on intrinsic motivation ( β = 0.236, CR = 4.420, SE = 0.053, p = 0.000). The higher the learner autonomy, the higher the intrinsic motivation. This phenomenon could be happened because the learner The test results indicated that learner autonomy had a positive and significant influence on intrinsic motivation ( β = 0.236, CR = 4.420, SE = 0.053, p = 0.000). The higher the learner autonomy, the higher the intrinsic motivation. This phenomenon could be happened because the learner
Thus, the research finding supports the hypothesis 16 that describes
a positive and significant relationship between learner autonomy and intrinsic motivation. Theoretically, this research also provides support for the previous research that hypothesized the higher the learner autonomy, the higher the intrinsic motivation as proposed by Standage et. al. (2005).
Through Structural Equation Modeling analysis, there is possibility to find the better model. This analysis is aimed to find the model that is stronger based on theoretical aspect. To figure the better model, the direct-effect model and fully- mediated model are constructed as alternative models, then they are compared to the partially-mediated model (initial model).
Based on Morgan and Hunt (1994), the three models (fully-mediated model, partially-mediated model, and direct-effect-only model) are examined in SEM model, then compared based on four different criteria:
a. Overall model fit as measured by CFI, CFI are index value that can be