Judy Sheard, Martin Dick, Selby Markham

March 2001

This is Part 2 of a report on a study that aims to determine attitudes towards questionable work practices of undergraduate students within the School of Computer Science and Software Engineering (CSSE).

More details of the study and the research method used may be found in Part 1 of this report at ???

The students’ ratings of the acceptability of the scenarios were analysed using a factor analysis. This is a method used to determine a latent variable structure that could account for intercorrelations of an observed set of variables. The factor analysis method performed used a Principal Axis Factoring extraction with a Varimax rotation with Kaiser normalization.

The initial factor analysis yielded three factors with eigenvalues greater than 1.0. Examination of the eigenvalues showed the fourth eigenvalue had a value of 0.957, just below 1.0. To determine if it was reasonable to extract a fourth factor from these results a scree test was used. A scree test is performed by plotting a graph of eingenvalues against the factor numbers. The point at which the curve of the graph levels out indicates the number of true factors which are present in the data. The scree test graph produced in this case shows a levelling out occuring at the third or fourth factor indicating a fourth factor is reasonable. The factor analysis was then repeated to produce four factors.

Comparison of the the rotated factor matrices produced for the three and four factor solutions show clearer results for the four factor solution. Examination of the variable loadings for this solution using a minimum variable loading of |0.4|, indicate interpretable results for each factor. The variable loadings are shown in Table 1and the scenarios within each factor structure are shown in Table 2.

Table 1 Rotated Factor Matrix (rotation converged in 6 iterations)

The students’ ratings of the likelihood of each reason causing cheating were analysed using a factor analysis. The factor analysis method performed used a Principal Axis Factoring extraction and a Varimax rotation with Kaiser normalization.

The factor analysis yielded three factors with eigenvalues greater than 1.0. The factor analysis yielded three factors with eigenvalues greater than 1.0. Examination of the variable loadings within the rotated factor matrix produced indicate interpretable results for each factor. The variable loadings are shown in Table 3 and the reasons within each factor structure are shown in Table 4.

Table 3 Rotated Factor Matrix (rotation converged in 6 iterations)

The students’ ratings of the likelihood of each reason preventing cheating were analysed using a factor analysis.

The factor analysis method performed used a Principal Axis Factoring extraction and a Varimax rotation with Kaiser normalization.The factor analysis yielded three factors with eigenvalues greater than 1.0. Examination of the variable loadings within the rotated factor matrix produced indicate interpretable results for each factor. The variable loadings are shown in Table 5and the reasons within each factor structure are shown in Table 6.

Table 5 Rotated Factor Matrix (rotation converged in 6 iterations)

The relationship between the students’ practice of a scenario and their ratings of the acceptability of a scenario were determined using Pearson’s correlations. These were performed for each scenario. A strong relationship (significant at the 0.01 level) was shown for each scenario. The strongest relationships were shown for scenarios involving assignment work. These were as follows:

• Scenario 10 (r = 0.50) This involved taking an assignment from a lecturer’s pigeon and copying it. The students also rated this the most unacceptable practice.
• Scenarios 11 (r = 0.41) and 17 (r = 0.45). These involved plagiarism from a Web site or a book without and not from the work of other students.
• Scenarios 1 (r = 0.45) and 5 (r = 0.40). These involved submitting part or all of a friend’s assignment work.

Table 7 shows the percentages of students who practised each scenario and the percentages of those students who knew someone who had also practised the scenario. For most of the scenarios a high percentage of students who admitted to practising the scenario knew someone else who had also practised the same scenario. The lowest percentages are for the scenarios that were also rated the most unacceptable.

The means of the ratings of the likelihood of each reason causing cheating were calculated for students who had practised a scenario and also rated it as unacceptable (> 3 on the Likert scale). This was calculated for each scenario.

The strongest reasons most commonly stated were:

• Not enough time
• Too great a workload at university
• Will fail otherwise

These reasons also comprised the third factor in the factor analysis of the cheating reasons.

The frequency of cheating practice is shown in Table 8. The numbers and percentages of students who have practised each scenario are shown. Scenarios 2 and 3, which are not considered cheating practises, have not been included in these totals.

The students were classified into groups according to their ratings of acceptability of cheating practices. This was done using a cluster analysis which is a method used to classify cases into groups based on a specified set of variables. The cluster analysis method performed was K-Means Cluster using the scenario ratings as variables and specifying three clusters.

The three groups produced by the cluster analysis will be referred to as low (n=227), medium (n=165) and high acceptability of cheating (n=65) groups.

An ANOVA was used to compare the differences between the groups on the acceptability of the cheating practices. There were significant differences for each scenario.

An ANOVA was used to compare the differences between the groups on the likelihood of the reasons causing cheating (Question 22). There were significant differences for each reason. For the high acceptability of cheating group all reasons except not enough time or too great a workload were more likely to cause cheating than the for the other groups.

An ANOVA was used to compare the differences between the groups on the likelihood of the reasons preventing cheating (Question 23). There were significant differences for five reasons:

• want to know what your work is worth
• pride in your work
• can get good marks without cheating
• against your moral values
• fairness to other students

For the high acceptability of cheating group all reasons were less likely to cause cheating than the for the other groups.

Cross tabulations were performed on the cluster groups against students classified according to campus of study, year level, study mode (full or part time), fee status (full fee paying or HECS), gender, age group (younger or older), or average course performance to date (low or high).

These showed that there are significantly more full fee paying students and Caulfield campus students in the high cheating acceptability group.

The students were classified into groups according to their admissions of cheating practices. The cluster analysis method performed was K-Means Cluster using the scenario ratings as variables and specifying three clusters.

The three groups produced by the cluster analysis will be referred to low (n=86), medium (n=268) and high cheating practice (n=95) groups.

An ANOVA was used to compare the differences between the groups on the acceptability of the cheating practices. There were significant differences for each scenario, except for the practice of posting to the Internet for assistance.

An ANOVA was used to compare the differences between the groups on the likelihood of the reasons causing cheating (Question 22). There were significant differences for each reason. For the high cheating practise group all reasons except not enough time or too great a workload were more likely to cause cheating than the for the other groups.

An ANOVA was used to compare the differences between the groups on the likelihood of the reasons preventing cheating (Question 23). There were significant differences for six reasons:

• want to know what your work is worth
• pride in your work
• can get good marks without cheating
• against your moral values
• never thought about it
• fairness to other students

For the high cheating practice group all reasons were less likely to prevent cheating than for the other groups.

Cross tabulations were performed on the cluster groups against students classified according to campus of study, year level, study mode (full or part time), fee status (full fee paying or HECS), gender, age group (younger or older), or average course performance to date (low or high).

These showed that there are significantly more full fee paying students and Caulfield campus students in the high cheating practice group.