This chapter describes and rationalizes the methodology that leads the study in order to answer the stated research question, namely revealing the way students perceive academic dishonesty at the university where the research was carried out.
The methodology is the section of research papers that serves as a ‘roadmap’ and tells what methods were chosen and how the research was conducted. What is more important, the section provides a justification for methods of data collection and analysis and gives the opportunity for further “…analysis, critique, replication, repetition” (Given, 2008, p.516). The given chapter includes the following sections: research design, research site, sample, research methods, data analysis approach, ethical considerations and limitations.
Research design section is the first section that explains the choice of quantitative research design to answer the research questions. It is preceded by a sample description, which describes the criteria for selecting the respondents, and the research method section which aims to justify the method used and demonstrate how the research was conducted. Finally, the chapter presents a data analysis section that lays out the step by step procedure applied for analyzing the obtained data, ethical considerations which were followed over the research period and
limitations of the study.
Research design
The research design section describes and justifies the methods that are used to hold the study. This research is built upon the quantitative design. Quantitative research is non-
experimental or as it is also called “ex post facto research” as the researcher in this type of research does not have any control over the independent variables (Hoy, 2010, p.17). In other words, the number of such independent variables as gender, age, major, year of study, GPA,
tuition form, and social status in the quantitative part of the research is not random and
controlled. In addition, non-experimental quantitative research allows the researcher to study the relationship between the independent variables and dependent variables in a real-life context (Mujis, 2011).
The purpose of the given research is to analyze students’ perceptions of academic dishonesty at a code-university by means of revealing the relationship between students’
attitudes towards academic dishonesty and their probable behavior. It also aims to define the factors that affect students’ attitudes and behavior. The research attempts to cover as many respondents as possible to see the overall trend at the university, so quantitative design has opted since this research design helps the researcher “establish the overall tendency of responses from individuals and to note how this tendency varies among people” (Creswell, 2014, p. 13).
Data collection was carried out by means of the cross-sectional survey that tests the hypotheses and helps to generalize the results of many samples (Hoy, 2010). The survey was constructed on the basis of the analysis of previous literature on this topic.
Research site
The research site was the most challenging part of the given research. To begin, it was supposed to be a comparative analysis of two universities with academic integrity policy and without. However, if a year ago it was not difficult to find a university without an academic integrity policy, today it is practically impossible due to the fact that universities, institutions, colleges, and schools have started to introduce the integrity policies into their academic process, which I find a positive step towards enhancing the quality of education. The second challenge was the research site access. I wrote to two universities regarding research site access, but was rejected by one university and did not get a reply from the another one. It might have happened
because of the sensitivity of the research topic. Due to the challenges the research was narrowed down to a case study and was taken in one Kazakhstani university. This university is not large in terms of undergraduate academic programs it offers. It trains specialists in various fields within both full-time and distance education formats. The programs are taught in two languages, namely Kazakh and Russian, but some language courses are taught in English. The academic programs of the university are accredited and every year the university is positively ranked by the MES RK and The National Chamber of Entrepreneurs of the Republic of Kazakhstan “Atameken”.
The university suited well my research purpose since it has the official academic integrity policy which was introduced in 2018 and published on the official website for public access.
While acquiring access to the university, I was interviewed by the Vice-Rector for Academic Issues on the purpose of my research. In addition to a letter of support that I gave, I was asked to show my survey to make sure that the questions are relevant and appropriate for the university population.
Sample
During the data collection process, the given quantitative research relied upon convenience sampling because the sample was going to be nonrandom. It means that every volunteer student (except first-year students and part-time ones) could participate in the survey (Johnson & Christensen, 2012). The survey was conducted among undergraduate students and included about 7% of the total undergraduate population to reach more accuracy and make better generalizations of the results for the total population. However, according to Johnson and
Christensen (2012), it is difficult to generalize from convenience sampling due to the fact that researchers may not know what group of population the individuals represent. The following independent variables were accounted for: gender, age, academic program, year of study, GPA
and tuition forms. As the sample was convenient, which means that all cases were taken into consideration (Given, 2008, p. 800), the number of males and females, students with the same GPA, tuition form and other characteristics was not equal. The research did not include first-year students, as at the time of the survey they did not have enough academic experience for the research because the academic period was not finished and they did not have their own strong perception of academic dishonesty. Another group of students which was excluded was distance education students because the survey was allowed to be taken on paper, which in its turn limited access to students.
Data collection
In this section I will describe the research methods that were used for the study. When I received the Ethics Committee approval to carry out the research, I started to look for the research site. After some failures, I gained accessed to one of the universities and explained the purpose of my research and how I wanted to hold it.
As aforementioned, for quantitative data collection, I applied a cross-sectional survey that was developed in accordance with the literature review. A cross-sectional survey was chosen due to its convenience since it provided quick and one-off access to a population (Creswell, 2014, p.
404). The survey was developed on the basis of the present surveys designed by O’Neill and Pfeiffer (2012), von Dran et al. (2001), Bisping et al. (2008) and in accordance with the literature review.The survey included two sections. In the first section, students provided their background information (independent variables) such as gender, age, academic program, year of study, GPA, and tuition form. In the second section, students were given the statements with which they expressed their agreement in accordance with the five-point Likert scale: strongly disagree, somewhat disagree, neither agree nor disagree, somewhat agree and strongly agree. When the
survey design was finished in Qualtrics, I placed the link on my Instagram account and asked my followers, who came from academic environments or student communities, to take it. I found this way to be the fastest to contact those who could suit my focus group. First of all, the participation was voluntary and, secondly, when I looked through the responses I could not identify them. In addition, I asked my peer-students from the master’s cohort to take the survey too. My focus group helped me to check the content validity of the statements. Then I checked the reliability of the survey in SPSS to see the internal consistency. Cronbach’s Alpha coefficient was equal to .884 which demonstrated the internal consistency of the test.
The online survey was rejected by the research site gatekeeper, so I prepared the paper survey. The vice-rector assigned me the person who looked through the schedules and took me to the classrooms. After that, I explained to the students who I was and why I attended their
university. Then, the purpose of research and research procedure were explained. I gave them the survey and a copy of a consent form so that they could take it with them. Students completed the survey in the classroom. When they finished the survey, they gave it back to me. Survey
responses were not shown to the administration, so I could ensure students’ confidentiality. I managed to cover not all students despite the gatekeeper’s permission the access to some
faculties was rejected. In total, 196 participated in the survey which comprised around 8% of the total population of about 2400 students. When the survey was completed, the data were put into the Qualtrics for further analysis.
Data analysis
When I inserted the data into Qualtrics, I downloaded them to analyze with Statistical Package for Social Sciences (hereinafter referred to as SPSS). The choice for SPSS was obvious because, first of all, this software is widely used for research in social sciences (Mujis, 2011)
and, secondly, its availability and easy access at Nazarbayev University support my choice for this statistical tool.
The first step for data analysis was data cleaning, which I did in the data view section. I looked through all responses and deleted those responses which were empty for more than one question. First of all, students could skip the questions accidentally when they answered the survey, but if there were responses with more empty values it could happen on purpose.
Secondly, the deletion of responses with one empty value could reduce the sample size which could negatively affect the analysis. As for the necessity of data cleaning, this process is necessary as it improves the quality of data and makes them more consistent.
The initial analysis that I conducted was a univariate analysis, which allowed me to make a descriptive analysis of the values to prepare the participants characteristics. The univariate analysis also helped to detect the outliers in the data set. After the detection of outliers, it was decided to either change the code or not to drop them. Two variables underwent recoding: a form of study and a language of instruction. For the former, 1 student out of 156 noted that he or she had a rector’s grant, so the category 3 for ‘other’ was changed into 2 for ‘state grant’. For the latter, 2 students noted that their language of instruction was English, but in fact the university provided education only in two languages. Therefore, the decision was made to recode the 1st response from category 3 (other) to category 2 (Russian) and the 2nd one to category 1 (Kazakh).
The outliers for dependent variables presented in the form of Likert scale were not dropped since, first of all, the initial responses were reduced to 156 and data drop could influence the further analysis, and, secondly, it was difficult to predict if the response was an outlier or students’ answer.
After the univariate analysis, I separated all tests depending on my research questions.
First, I conducted the ordinal logistic regression or in other words polytomous universal model (hereinafter referred to as PLUM) to look at the relationships between more than two nominal or ordinal variables (Mujis, 2011) or in my case to check if independent variables predicted
students’ perception of the effectiveness of academic integrity policy. The next ordinal regression tests were conducted to see if independent variables affected students’ attitude to academic violations and the likelihood of their involvement in these violations. Along with causation tests, I held the correlation analysis to reveal the relationships between students’
attitudes to the academic violation and the degree of the likelihood of their involvement in them.
For that reason, Spearman’s Rho coefficient was calculated to reveal the statistical significance in the relationships between two ordinal variables, and the effect size was defined to look at
“how strong the relationship is” (Mujis, 2011, p. 109).
In addition, before ordinal logistic regression, I prepared the variables for analysis. Two research questions aimed to determine the factors or independent variables that influenced students’ attitudes and behavior. The given independent variables were represented by nominal variables, such as gender, program, the language of instruction, tuition form and ordinal
variables, such as enrollment year and GPA. According to Mujis (2011), as the nominal variables are not ordered, dummy variables should be created before regression analysis. Therefore, the
‘males’ variable for gender was turned into a reference category, ‘Russian’ – for language of instruction, ‘state grant’ for a form of study and ‘other’ for program.
The independent variables were also checked for multicollinearity in order to avoid misleading results. To check the multicollinearity, independent variables were correlated and for further ordinal regression the variables that were not correlated or poorly correlated with each
other were chosen. For that purpose, the multicollinearity matrix demonstrated by Kenton (2020) was calculated (Table 1) that helped pick up the set of the most appropriate independent
variables.
Table 1
Multicollinearity Matrix
Variable Gender Program Form of study
Language of Instruction
Enrollment Year
GPA
Gender -
Program .000 -
Form of study
.061
.234 -
Language of Instruction
.000
.000 .141 -
Enrollment Year
.119 .028 .042 .432 -
GPA .058 .813 .187 .200 .213 -
This table does not demonstrate a strong correlation between any of the variables, but shows a moderate correlation between enrollment year and language of instruction. It also could be seen that for regression analysis the set of gender, language of instruction and form of study could be taken or, for example, from of study and language of instruction.
Ethical considerations
Prior to data collection, I passed ethics training at Nazarbayev Univerity and sent my application with all accompanying documents for ethics approval to GSE IREC (Institutional Research Ethics Committee). As the survey was not intended for first-year students, nobody under 18 could participate in the survey. Therefore, the research was characterized as no more than minimal risk.
The data collection did not start until the access was approved by the gatekeeper on behalf of the university vice-rector. I showed the letter of support and my survey and answered
the questions regarding the data collection procedure and publications. I also explained how I would ensure the confidentiality of the university name and students who would participate in the interview. In order to ensure the confidentiality of the university, the programs were coded.
During the survey data collection, students were explained the purpose and benefits of the research as well as the right not to participate if they do not want or return the empty survey. All of them were given the copy of consent form.
Limitations
The given research has some limitation that I will describe in this section. First, the number of respondents might not comprise 10% of the total population. At this university, there are about 2400 full-time undergraduates in the second, third and fourth years of study but the study analyses only 156 responses. In fact, there were more than 190 participants but after the data were cleaned, the number was slightly reduced to 156 which is only 7% of the total
population. The small sample size could affect the generalizability of the research and make the findings not representative for the population.
Second, the research does not present a wide range of programs. In fact, only four of them participated due to access issues and students’ availability.
Third, the number of males and females in the research is not equal that is why probably it would influence the research findings in terms of considering gender as one of the factors that affect students’ perception of academic dishonesty.
Last but not least, the survey was built on the five-point Likert scale which made it possible that participants did not take it seriously. While conducting the research, I noticed that for some students it took only about 5 minutes to answer the questions; that is why the
probability of random answers cannot be denied.
Chapter 4: Findings of the study