Statistical analysis is performed to compare the variance and frequency of dimensions/factors that may affect academic integrity.
Variability in country results
Analysis on single and multiple answers survey questions is performed to get a better understanding of which of the three countries is more affected with cheating and plagiarism factors.
Single answer questions
In order to test the effect of the three countries on the cheating and plagiarism factors, an analysis of variance was conducted on all the questions and reports the F statistic along with p-value. The significant results were reported in terms of the Krushal F statistic and p-value with significance level of 0.05.
To get a deeper understanding of which pairs of the three countries is affected by each of the significant cheating factors (as defined in section 2), we performed the multiple comparisons of means using nonparametric test entitled “Dunn’s test”. This test does not assume the data is coming from particular distribution.
We test the three following null hypotheses:
-
H0: The median of a factor/dimension that can influence academic integrity in Egypt has same distribution as in KSA.
-
H0: The median of a factor/dimension that can influence academic integrity in Egypt has same distribution as in Jordon.
-
H0: The median of a factor/dimension that can influence academic integrity in KSA has same distribution as in Jordon.
Where factors and dimensions are as identified in section 1.
Table 2 shows the results of this test in terms of p-values. Rejecting (True or False) the null hypothesis is represented in the last column entitled ‘reject’ of Table 2, where family-wise error rate was adjusted by using Bonferroni correction.
Egypt and KSA
Results indicate that at 0.05 level of significance, cheating on an exam factors including, gender (Q7), level of education (Q10), and years of study (Q20) [first factor], pressure from parents (Q8), collectivism (Q12) versus individualism, social (Q17) [second factor], academic reasons (Q19), lack of teachers’ control (Q14) [fourth factor], the use of electronic devices when not allowed in a test (Q2), copying partially or completely another person materials (Q6) [forms of cheating] have different means of distributions in Egypt and KSA.
Egypt and Jordon
There is no evidence at this level of significance that the distribution of cheating on an exam and test via gender (Q7) [first factor], pressure from parents (Q8) [second factor], electronic devices (Q2), academic reasons (Q19) [fourth factor] are different between Egypt and Jordan copying completely or partially another person work (Q5, 6) [styles of cheating].
Jordon and KSA
Results indicate that at 0.05 level of significance, cheating on an exam factors including, gender (Q7), level of education (Q10), and years of study (Q20) [first factor], pressure from parents (Q8) [second factor], academic reasons (Q19) [fourth factor], copying partially or completely another person materials (Q5,6) [styles of cheating], have different means of distributions in Jordon and KSA.
Note that years of study (Q20) and levels of education (Q10) distribution are different between any pairs of the countries at 0.05 level of significance.
Compared to Egypt and Jordon, KSA maintains advanced and robust Internet and mobile phones digital infrastructure (CTIC, 2021). Meanwhile, average annual income is higher in KSA compared to Egypt and Jordon (GASTAT, 2021), thereby KSA students have an access to high-end mobile phones, laptops, and wireless electronics compared to their peers in Egypt and Jordon. Moreover, students in schools and higher education institutes have a reliable access to the Internet in their classrooms. Jordon per capita income (~ $ 4282) is a little higher than Egypt (~$3547) (WBG, the World Bank Group, 2021), however students in Jordon still do not have an access to a reliable digital networks in schools and some public educational institutes. This limited access to high-end digital networks make it harder for students in Egypt and Jordon to cheat using electronic devices compared to KSA. Additionally, booming population, low national income, and high unemployment rates (WBG, the World Bank Group, 2021) make it is very competitive for students to secure a place in the Egyptian higher education system. For the same reason parents put high pressure on children to earn high marks in order to secure admittance in the Egyptian universities, which make students cheat to get high marks. Large number of students per class in Egypt limits teachers’ ability to control cheating compared to KSA and Jordon. Egypt and Jordon are ahead of KSA in giving females opportunities for education and work. With KSA started to place women empowerment as one of its national’s priorities (Smith 2020), KSA give females more educational opportunities and females try to prove themselves by earning higher marks than males. This competitive environment can lead to cheating from both genders.
Multiple answer questions
The survey contains two multiple-answer questions addressing the main reasons of cheating and plagiarism according to the students from different countries. The proportion distribution of these main reasons is shown in Figs. 2 and 3. In all three countries, ~ 50–60% of the students cheat Q (16) because of laziness [third factor]. However, a student’s main reasons for misbehaving are to not fail in a test or exam. Moreover, the majority of KSA students (~ 70%) care about grades more than Egyptian (~ 47%) and Jordanian (~ 56%) students.
An interesting difference occurs in the distribution of whether a student cheats because of their jobs. Less than 10% of Egyptian students blame their misbehaving on having another job compared to 20% in KSA and 30% in Jordan.
As for plagiarism Q (18), ~ 50% of the KSA student sample use plagiarism due to laziness and ~ 55% care about grades. This percentage in KSA students is significantly higher than Egypt and Jordan (~ 40% - 45%). Only 15% of Egypt’s students considered writing on their own is hard as opposed to Jordan and KSA students (~ 30%). A majority of Egyptian student sample consider their laziness and willingness to get high grades encourage to plagiarize. Around 55% of the students in KSA expressed not having the ability or the information on the assignment compared to 64% in Egypt and 61% in Jordan.
It is interesting that students in the three countries consider laziness as one of the main reasons for cheating or for plagiarizing their work (Fig. 2). Many of the students do not want to put enough time to study. Reluctant to study could be because students are either not interested in a specific subject or in their major. This is common in many Egyptian universities, where students are forced to pick a specific major based solely on their marks in one exam (high school national exam) that they have to take at the end of grade 12.
It is interesting to find out (Fig. 3) that only a small percentage of the students think that the length and difficulty of the material is not the major reason that make students cheat.
Dominant exam misbehavior factors
Across all the observations of the three countries, 110 observations have missing values. In order to reduce bias and errors in applying any of the imputation techniques, a decision was made to remove those observations from the analysis.
To further explain, the differences we observed and reported above, we examined the roles of different factors in cheating and plagiarism behavior.
The first group of factors of cheating included individual characteristics such as gender, years of studying. The second group included educational context such as level of education, and rate of academic performance. The third group included motivational and contextual factors like neutralization attitudes, classroom environment and so forth.
Egypt, Jordan, and KSA were compared separately. Different feature selection methods were implemented in order to get the subset of features that best explains the variation in cheating factors.
These statistical methods are Recursive Feature Elimination (RFE), Linear Regression, Lasso and Ridge, and Random Forest. We then rank the features using the mean rank ordering of all methods. Below is a brief explanation of each method.
Feature selection is one of the important steps to apply in model building. It enables us to rank the features that explain most of the variability observed in the outcome. Different methods are then implemented to reduce the number of features in the final model. The methods RFE, Lasso, Random Forest, and Ridge regression used in this analysis are some of the robust methods used to trim down the number of features. For this analysis, there is no particular reason on why we choose one method over the other. All four methods were implemented to show that the final features were chosen carefully by applying and averaging many validated techniques for measuring features’ importance. The assumptions tested are the ones specific for building linear regression models and testing the importance of each feature using the t test.
Recursive Feature Elimination or RFE uses a model (e.g. linear Regression or SVM) to select either the best or worst-performing feature, and then excludes this feature. The whole process is then iterated until all features in the dataset are used up (or up to a user-defined limit).
Lasso
This method picks out the top performing features, while forcing other features to be close to zero. It is useful when reducing the number of features is required, but not necessarily for data interpretation, since it might erroneously indicate that some features do not have a strong relationship with the output variable.
Random forest
This is an impurity based ranking that is typically aggressive in the sense that there is a sharp drop-off of scores after the first few top ones (whereas for the other ranking methods, the drop-off is clearly not that aggressive).
Ridge regression
This method forces regression coefficients to spread out similarly between correlated variables.
The mean rank obtained by applying all methods of feature selection along with final parameter estimates for each country (with confidence interval limits) is represented in the results. A stepwise selection excluding all non-significant features at 0.05 level of significance has been applied on the top of features selection.
Egypt analysis results
On the left panel of Fig. 4a, one can see the top three features are related to individuals and academic characteristics. These features are academic performance (Q11), gender (Q7), and level of education (Q10) (first factor). Note that lack of teach control (Q4 and Q14) [fourth factor] was excluded because of its insignificance at a 0.05 significance level when performing a t-test which is testing the null hypothesis (H0: B = 0).
By fitting multiple linear regression, 79% of the variation in cheating is explained by the top three factors.
Given the rate of academic performance and level of education are constant, males increase the estimated rate of cheating by 1.27 on average. Furthermore, given the gender and the level of education are fixed, on average.
Given the gender (Q7) and academic performance (Q11) are fixed, on average, every unit increase in level of education (Going from Graduate to undergraduate), the estimated unit of cheating increased by 0.43.
KSA analysis results
Applying feature selection to the KSA Country, results are very similar to those of the Egypt sample analysis.
The top three selected features selected are academic performance (Q11), gender (Q7), and level of education (Q10) (first factor). Note that extrinsic motivation (Q13) [third factor] is excluded because of its insignificance at level of 0.05 when performing t-test for testing the null hypothesis (H0: B = 0) (Fig. 4b).
Similar results to Egypt with change in values for parameter estimates appear in KSA. Given the rate of academic performance and level of education are constant, males increase the estimated rate of cheating by 3 on average. This result approximately is doubled what we saw in Egypt.
Given the gender and the level of education (Q10) are fixed.
Given the gender and the academic performance (Q11) rate are fixed, on average, every unit increase in level of education (Q10) (Going from Graduate to undergraduate), the estimated unit of cheating increased by 0.25. This effect is less than Egypt.
Jordan analysis results
For Jordan, educational and contextual characteristics play big roles as factors for cheating. As shown in Fig. 4c, unlike Egypt and KSA, the most important features are lack of teach control (Q14) [fourth factor], academic performance (Q11) [first factor], neutralization attitudes (Q15) [third factor], and extrinsic motivation (Q13) [third factor]. By applying stepwise selection, on average, 83% of estimated variation in cheating is explained by neutralization attitudes (Q15) and academic performance (Q11).
Given a fixed rate of performance academic, the increase unit in neutralizating attitudes (Q15) (a student can’t finish his education without copying someone else work) will increase the estimated cheating by 2 units. Approximately, on average, same increase in unit of cheating is observed if rate of academic performance decreases by one rate with “Neutralizating attitude” being fixed.