Thesis and Publication Tips for Epoka University Graduate Students in Business Administration
Teoman Duman, Ph.D.
The following are handy tips for graduate students of the Business Administration Program at Epoka University. They will help students understand the thesis/graduate project preparation process.
Thesis/dissertation/graduate project are some of the categorizations of “graduate final work„ in different universities across the world. Categorizations usually differ according to program requirements. At Epoka University Business Administration graduate program, we have two kinds of final work:
Final work with a professional nature and usually investigates real-world sector/business/organizational problems. Students who conduct professional research try to understand the nature of sector/business/organizational processes through case studies, benchmarking research, and related context-focused analyses. These studies usually use exploratory and descriptive research methodologies. The focus in these studies is on the implications to the sectors/businesses, which arise from the research findings. These studies are not usually concerned with developing research variables and testing hypotheses. They are more concerned with certain research questions that try to answer how things happen rather than why they happen.
Final work with an academic nature, on the other hand, is more explanatory in content trying to understand why certain things happen. All necessary research methodologies (i.e. exploratory, descriptive, explanatory) are used to identify variables and test relationships among them. In these studies, relevant independent (cause) and dependent (resulting behaviors/occurrences) variables should be identified and situational effects should be analyzed to explain relationships in more detail.
The following discussions will clarify the category of research studies mentioned above. The first step for a graduate student is to identify the nature of the graduate work he/she needs to conduct (professional vs academic). In both studies, data are needed, but the data needs will depend on the research questions and/or hypotheses established before starting the research.
In the Epoka Business Program, we have three categories of final work:
1. Master's projects – professional study in nature
2. Master's theses – academic study in nature
3. Doctoral theses – academic study in nature
Finding a research topic
Most graduate students get confused when they have to decide on a research topic. The golden rule is that at least half of the job is finished when research questions (and hypotheses) are clear in graduate research. Research questions and hypotheses are so important that they serve as the guide for the research study. Three key points to remember when searching for a topic are:
1. Is there a current topic that you have already worked on during your undergraduate studies? Do you want to continue working on such a topic that you already have spent some time on?
2. Is there a topic that is relevant to your future career needs, your personal goals (i.e. family businesses, new investments?), your expectations and desires? How can you use the information base you will develop by doing graduate research later in your life?
3. Is there a topic that is becoming popular in social/business life and promising a future for upcoming generations? Where is the world of business going now and how can you learn about more what is coming in the near future, which will help you to prepare yourself for future challenges? If you don't have an idea on this issue, can you search and learn what it can be?
In research, the “million-dollar question“ is always the “why“ question. For what purpose will your research serve? Who needs your research? Who will benefit from your findings? Why are you doing this research?
The first step in deciding on a research topic (and relevant research questions and hypotheses) is a well-done literature review. A good collection of relevant literature sources, a comprehensive reading of these sources, and a good understanding of the discussions in the literature will bring you to the point of deciding on research questions and hypotheses. To cook ideas and get different perspectives from others, discussions with friends, colleagues, and professors are necessary. Said differently, after a careful literature review and discussions with people around you, especially with your professors and your advisor, you will come up with a research topic and relevant research questions and hypotheses. When you are ready with your research questions and hypotheses, you can start devising your research methodology.
Today, the internet is a perfect source where students can find many popular topics. Not only can students reach many different literature sources, but also they can benefit from ideas on the discussion forums.
Tips about Professional Master Projects
Master projects are short-term research studies where students analyze a topic that will enhance their knowledge of the topic as a supplement to their professional master's education. Some master projects may also benefit to industry professionals if the topic is interesting enough and the research study bears contemporary solutions to ongoing problems. Such creative master projects are rare. But, everything depends on how interested the student is and how he/she is serious to bring out a valuable project. Just as in any academic research, harmony between the student and the adviser will identify how productive these research studies will be.
Master projects usually cover the same credit value as one or two master courses and the student is expected to write a project report bringing together all literature evidence and research findings related to the topic. In these projects, students are not supposed to conduct full-scale empirical research, but they can conduct some field research to collect data to find answers to research questions. The extent of the research project is between the student and the advisor. However, as a general rule, the scope of the study should not extend a student’s education period in the master's study. As per the topics, most master projects deal with sectoral problems, analysis of current business applications, case studies of exemplary organizations, etc. At Epoka University Business Program, the professional master project is considered as a semester course with 9 ECTS credit value.
Tips about Master Theses
Master theses usually take one year to complete. There may be different implementations of master theses processes in terms of credit hours, semester periods, etc. in different parts of the world. But, for a settled master's thesis, one year is a reasonable period. A master's thesis can be considered as an earlier step of doctoral study. Therefore, master theses should lay the ground for doctoral studies.
In master theses, we usually require students to conduct an empirical research study in which students will collect some primary data and analyze them as their own contribution to that scientific area. This shouldn’t sound like a full-scale doctoral study, but it should mean that if a student is spending one year of his/her time on a topic to conduct research, then he/she should present some new findings that will sound interesting to everyone. How this can be is up to the student and the advisor, of course. Most master theses are average but some master theses are as good as they can be. Most master studies collect primary data from organizations, consumers, the general public, etc. and utilize exploratory or descriptive methodologies in data collection. But, the general expectation from a good master thesis is to use explanatory research methodologies and try to analyze cause-effect relationships. Models tested in master theses are usually simple and take one aspect of an extended research model. However, they may be very informative if the model is set up well, grounded in theory, and the research is reliable and valid to test the model scientifically.
Tips about Doctoral Theses
Doctoral theses, of course, are the highest level academic research works and they should deserve this level of competency in their installation. Guidelines are much more established as compared to master studies. A good doctoral research study makes some special contribution to its scientific area and should be referenced well by other researchers. I advise my doctoral students first to think about continuing their research agenda from their master studies if they did a good job in their master's research and if they like to continue working on that topic. Most doctoral students have a hard time finding an attractive research topic that will make some special contribution to science, and quite some of them even drop out from doctoral studies because they inadequately propose a promising model that will pass the monitoring committee (By the way, this may sound scary but it is also true that most doctoral students have a hard time to construct such a valuable research framework and complete in two to three – sometimes even more – years.).
Before identifying a research topic and later a research model, the student should do a comprehensive reading about the topic and the theory, especially from strong academic literature sources. This topic should have both academic and practical aspects. In social sciences, this combination is important because every research in social sciences has some implication for both theory and real life. So, the topic should be interesting for both academic and professional communities. The student should try to answer two main questions himself/herself before deciding on a research topic: “What is my contribution to scientific inquiry with this topic” and “How will the research findings benefit practitioners in real life to develop/improve goods, services, processes, etc., which in the end will contribute to the betterment of individual and/or social life for humans?”. If the student can legitimize the topic in his/her mind and convince the advisor that the proposed topic is a valuable research topic, then, the student can continue doing some qualified academic reading on the topic.
My general rule is that the student should find (at least) one good doctoral thesis on the topic and read this thesis well to see how a typical doctoral thesis is prepared. Doctoral theses are very good sources of information on certain research areas. Sometimes, some well-prepared master theses also help to extend the student's knowledge of the topic. During reading, the student should take note of some strong, established publications about the topic looking into the references sections of those theses. Next, the student should collect all relevant conceptual and research articles from strong databases and start reading them to shape up his/her mind about the topic academically. In social sciences, most if not all students have to refer to different scientific areas to find relevant sources of theory and research. For example, most consumer behavior theories were developed by psychologists in the past. A good consumer behavior study is one that brings evidence from multiple scientific areas to show how the theory evolved in different disciplines. This is not a rule but more of an expectation for doctoral studies in social sciences.
Preparing the research proposal
The meaning of a research proposal may be misunderstood by many. A completed research proposal is the one that is ready to start data collection. Research committees should discuss the proposal and finalize their decisions on whether to give students a go-ahead for data collection based on the completed research proposal. Usually, a major part of the research process is finished if the proposal is ready. Following the proposal defense, a student can collect data and finalize writing the final work.
As I explained before, for master projects, preparing a proposal and then conducting the research may take equal time and can be finished in a semester or in a year altogether depending on the program requirements. Interested master's students who will do a master's study with a project can start talking about their topics during the first semester with their advisors and finish the work in the second semester. For masters with a thesis, students usually decide on their topic during the second semester and finalize their proposals at the beginning of the third semester. If their proposals are accepted then these students can start their master thesis research right after their proposal defenses, and they can finish their data collection and thesis writing at the end of the fourth semester. As I discussed before, the time frame and the systems of master and doctoral studies may differ in different parts of the world but the guidelines I mentioned here reflect the program requirements in most parts of the world.
I would like to conclude my discussion on the research topic and proposal here for now. But I also would like to put a final note on the literature sources. When my students ask me “Which publications/databases do you suggest for me to find literature sources for my study?”, I usually say; “there are two types of academic sources today “the good ones and the bad ones”. The good ones are the ones that publish good quality stuff but the bad ones publish everything. You should only use the good ones because they will reflect the quality of your research later”.
Choice between qualitative and quantitative methodology
Your research questions and hypotheses will identify the type of data collection methodology you will use in your research. There are two major data collection methodologies used in social sciences research.
Qualitative methodology: “Qualitative research is an unstructured, exploratory research methodology based on small samples that provide insights and understanding of the problem setting” (Malhotra, 2009, p.180). Qualitative methodology is used to identify the nature of the topic investigated in the research problem. Definitions and interpretations of the concepts are sought by a collection of qualitative data. This methodology is usually used to develop theory by identifying underlying reasons of questioned outcome variables. Sometimes this methodology is used to test behavioral theories. Researchers look for underlying patterns in the text data they collect from respondents to build their theories and models. Focus group interviews, in-depth interviews, and projective techniques are some examples of qualitative data collection techniques. Well-known software such as NVivo and Atlas.ti are used to analyze qualitative data.
Sample research with qualitative methodology - http://www.diva-portal.org/smash/get/diva2:428481/FULLTEXT02.pdf
Quantitative methodology: “Quantitative research methodology seeks to quantify the data and typically applies some form of statistical analysis.” (Malhotra, 2009, p.180). In quantitative research, researchers use numerical data to analyze relationships between coded variables. Most basic quantitative research is done for descriptive purposes. Numerical data are analyzed to describe the nature of the variables (i.e. frequencies, means, categorical analysis). In more complicated analyses, relationships among variables are analyzed to test models. Below, more detail on quantitative methodology will be given to show their usage in social science research.
Tips on Data Processing and Statistical Analysis Using Linear Modelling Techniques
In statistical methodology, two main statistical methods of data analysis are parametric and nonparametric. Parametric methods are based on probability theory operating on assumptions to estimate population parameters. Based on the probability theory and the central limit theorem, distributions of population parameters are expected to be approximately normal. Nonparametric methods on the other hand are distribution-free methods making no assumptions on the distribution of population parameters. Here, my tips will focus on parametric methods.
Common methods of data analysis in social sciences include (Agresti and Finlay, 1997; Agresti, 2018):
Descriptive statistics – Descriptive analysis of data includes analysis of frequency distributions (i.e. relative frequencies, histograms and bar graphs, stem and leaf plots), central tendencies (i.e. mean, median, mode, quartiles) and variation in data (i.e. the range, variance and standard deviation, internquatile range).
Sample research with descriptive analysis - https://sites.google.com/a/epoka.edu.al/teoman-duman/epoka/masters/3-4-bus-501-research-methods-in-business?authuser=0
Comparison of two groups – Means of two groups can be contrasted in a longitudinal or cross-sectional study where in the former one “a mean or proportion parameter are contrasted at two points in time„ whereas in the latter one “group means are compared in a single survey„ (Agresti and Finlay, 1997, p. 210).
Sample research with comparison of two groups - http://ro.ecu.edu.au/cgi/viewcontent.cgi?article=1016&context=smatl_pubs
Analyzing association between categorical variables – This methodology is used when both explanatory and response variables are measured with nominal or ordinal scales. To analyze associations between these variables, contingency tables are used. Associations among cell frequencies or proportions are measured with chi-squared (χ2), odds ratio and other relevant statistics.
Sample research with an analysis of the association between categorical variables - https://www.sciencedirect.com/science/article/pii/S2212571X16300129
Linear regression and correlation – This methodology is used to analyze the relationship between two variables measured with interval and/or ratio scales. The least squares method is used to measure the association between variables. The strength of the relationship is measured by the Pearson correlation formula. Sample research with linear regression and correlation - http://www.scielo.mec.pt/pdf/tms/v13n3/13n3a02.pdf
Multiple regression and correlation – This method is used to estimate variation in a response variable based on data from a collection of explanatory variables. By using this technique, a researcher can analyze partial and combined effects of independent variables on the dependent variable as well as possible interaction effects due to variation in different categories of independent variables.
Sample research with multiple regression and correlation - https://www.sciencedirect.com/science/article/pii/S2212571X17300136
Analysis of Variance – This technique is used to compare three or more categories of a dependent variable based on one or more measured independent variables. For example, a marketing researcher may want to find out perceptual differences among four variants of a product category based on consumer evaluations of product characteristics (e.g. taste, color, nutritional content).
Sample research with ANOVA - https://www.sciencedirect.com/science/article/pii/S0261517718300074
Multivariate statistical methodologies (more detail is given below) – These techniques are used to analyze relationships among multiple independent and dependent variables based on causality assumptions. Tests of mediation (i.e. chain relationships) and moderation (i.e. statistical interaction) are commonly used modeling techniques in multivariate statistical methodologies.
One popular area of research in social sciences is model building where social scientists try to analyze the causes and effects of behavioral variables. In linking causes and effects, data on cause and effect variables may show linear or nonlinear distributions. Commonly used linear-modeling-based data analysis techniques in behavioral sciences are listed below (Hair et. al., 2014). When the response variable is a single variable, the modeling method is called univariate statistical methodology whereas when more than one response variable is involved in data analysis, the modeling techniques are called multivariate statistical techniques.
Multivariate Statistics
Principal Component Analysis (PCA) - Exploratory Factor Analysis (EFA) – These methods are used to identify the dimensional structure of constructs. All measured variables are used in the data analysis tool to identify their associations with the estimated dimension(s).
Sample research with PCA and EFA - https://www.sciencedirect.com/science/article/abs/pii/0160738394901201
Multiple Discriminant Analysis – This method is used to identify group memberships of cases on a dependent variable (e.g. users – nonusers of a new product) based on certain independent variables (e.g. brand use experience, income level), which are used to identify discriminant functions (i.e. best model to discriminate group membership) to classify cases into dependent groups.
Sample research with multiple discriminant analysis - https://www.emeraldinsight.com/doi/pdfplus/10.1108/eb058393
Logistic Regression – This method is a linear modeling technique similar to multiple regression where the dependent variable is a dichotomous one. The presence or absence of a characteristic of the dependent variable (i.e. strong loyalty – weak loyalty for a product) is identified based on odds ratios identified based on logistic regression coefficients of a number of independent variables (e.g. brand awareness, user status, price perception, value perception).
Sample research with logistic regression - https://www.tandfonline.com/doi/abs/10.1300/J149v04n04_03
Conjoint Analysis – This technique is used to identify most influential features/functions/benefits of a product to estimate decision patterns of potential customers on a selected group of product alternatives.
Sample research with conjoint analysis - https://www.sciencedirect.com/science/article/pii/S0278431914000899
Cluster Analysis – This is an exploratory categorization technique where membership categories (i.e. clusters) and group memberships are identified based on measured variables in a research study. Each case is allocated in a cluster, which is identified based on the premise that group members share most common characteristics with each other (i.e. within group similarity).
Sample research with cluster analysis - https://www.tandfonline.com/doi/abs/10.1080/19368623.2011.536071
Multidimensional Scaling – Perceptual Mapping – This technique is used to identify similarities and differences among objects, places, brands etc. using indices created from user perceptions on attributes. Similarities and differences are identified on multidimensional space visually which help users to identify policies in a more practical way.
Sample research with multidimensional scaling - https://www.sciencedirect.com/science/article/abs/pii/0160738388900850
Analyzing Nominal Data with Correspondence Analysis – This research technique is used to analyze patterns among categorical variables presented in the form of contengency tables. In marketing, brand characteristics are frequently associated with consumer characteristics to identify brand choice patterns based on these associations.
Sample research with correspondence analysis - https://www.sciencedirect.com/science/article/pii/S0261517713000113
Confirmatory Factor Analysis (CFA) – This method is a multivariate linear modelling technique used to investigate construct validity of research concepts. It is commonly used to identify convergent and discriminant validity of research constructs, which set up the base for further structural model testing.
Sample research with CFA - https://www.sciencedirect.com/science/article/pii/S0261517714000399?via%3Dihub
Structural Equation Modelling (SEM) – This is a multivariate modeling technique where associations among multiple independent and dependent variables are identified based on construct covariances. The technique allows researchers to test complicated mediation and moderation hypotheses.
Sample research with SEM - https://www.sciencedirect.com/science/article/pii/S2212571X17300525
Multivariate Analysis of Variance (MANOVA) and GLM – This is a linear modeling technique where comparisons among categories of multiple variables are done based on a number of dependent response variables. Researchers set their models to test interactions in the models which identify contextual effects on the relationships.
Sample research with MANOVA - http://journals.sagepub.com/doi/abs/10.1177/1096348006297292
Analysis of Covariance (ANCOVA) – This technique is an extension of one-way analysis of variance where the relationship between a categorical explanatory variable and a metric dependent variable is controlled with quantitative covariate variable(s). Covariates are used to identify if the differences among categories of the independent variable change within different levels of the covariate(s).
Sample research with ANCOVA - http://journals.sagepub.com/doi/10.1177/004728759303100405
Profile Analysis – Similar to cluster analysis methodology, this technique aims to identify subject profiles based on observed data. The technique allows researchers to reduce the large scale of continuous and categorical data into smaller subgroups to extract hidden groups (i.e. segments) from measured variables.
Sample research with profile analysis - https://www.sciencedirect.com/science/article/pii/0143251681900578
Survival/Failure Analysis – This is a longitudinal research technique where the researcher can estimate the period during which an event can happen as well as the effects of certain covariates on the time distribution of events. Business researchers commonly use this technique to identify effects on a business's life span.
Sample research with survival/failure analysis - https://www.sciencedirect.com/science/article/pii/S0261517715300637
Canonical Correlation – This method is used to correlate two sets of variables and understand their relationships with each other. For example, consumers' perceptions of product attributes (e.g. quality, price, packaging, warranty terms) can be linked with shopping mode preferences (e.g. in store, telephone, website, mobile app) to see the level of correspondence between product attribute perceptions and shopping mode preferences.
Sample research with canonical correlation analysis - https://www.sciencedirect.com/science/article/pii/S0261517717300365
Multilevel Linear Modelling – This method is analogous to analysis of covariance in that it is used to analyze relationships between a set of independent variables and a dependent variable by adding further covariates to the model.
Sample research with multilevel linear modelling - https://www.sciencedirect.com/science/article/abs/pii/S2211973618300643
Multiway Frequency Analysis – Also known as loglinear analysis, this method allows researchers to analyze relationships among a set of discrete variables measured at nominal or ordinal levels.
Sample research with multiway frequency analysis - https://www.emeraldinsight.com/doi/abs/10.1108/09596110110388945
Time Series Analysis – Commonly used in economics and finance, this method allows researchers to analyze relationships between a set of explanatory and response variables to find out the effects of change in data in specified time periods.
Sample research with time series analysis - https://www.sciencedirect.com/science/article/pii/S0264999313003842?via%3Dihub
For more details, please visit: https://onlinecourses.science.psu.edu/stat505/node/1/
Data preparation
From beginning to end, a good research project is a unified whole composed of well-integrated parts. This integration is possible through meaningful construction of research proposals and the accuracy of data collection processes. Two concepts define the authenticity of an empirical research project:
Validity in measurement of variables: Variables are valid if they measure what they are supposed to measure in a correct way.
Reliability of the measured variables: Variables are reliable if they produce consistent results each time they are measured.
For validity, a proper scale development process is important. Scales should reflect the true nature of the variables each time they are measured with collected data. Content and meaning of the variables are the main concerns in validity.
For reliability, variables should be measured in an error-free way. Each time they are measured with certain scales, they should bring out consistent results.
A good researcher should “read the data” more than systematically analyze them. The researcher should detect all mistakes and inconsistencies in the data before running the analyses. Identifying mistakes and eliminating erroneous parts in the data are called data cleaning. In data cleaning, the researcher looks for incorrect entries, outliers, unusable cases variables due to missing data, etc.
Once, the researcher identifies and eliminates all inconsistencies and mistakes in the data, he/she can continue with testing the assumptions of the statistical methodology he/she will be using to test the research proposals or hypotheses.
Here, I will briefly describe the assumptions of linear model testing.
Analyzing linear model assumptions
Linearity assumption: Linear models are sensitive to nonlinear data. To analyze the linearity assumption, residual plots should be investigated. Standardized residuals and standardized predicted values can be plotted and checked for linear relationships.
Multivariate normality assumption: One important assumption of linear models is the normal distribution of the population parameters. For this, variables used in data analysis should stay within the limits of normality before they are used in the model testing. Researchers can analyze data plots to see if there is any visual discrepancy from normal distribution as well as analyze P-P plots with skewness and kurtosis values to see if they are in the range to assure normal distribution of the variables.
Lack of multicollinearity assumption: Multicollinearity occurs when correlations among independent variables are higher than desirable levels. Multicollinearity prevents researchers from identifying individual effects of independent variables on the dependent variables. Tolerance and VIF values can be checked for thresholds for desired correlation values among independent variables.
Homoscedasticity assumption: “Homoscedasticity means that the variance of errors is the same across all levels of the independent variable” (Osborne and Walers, 2002). Violation of this assumption implies heteroscedasticity in the variable. Plots of standardized residuals and standardized predicted values can be examined if these distributions show any undesired distributions visually (Osborne and Walers, 2002). Bartlett’s and Levene’s tests can be used to check if there are violations of equal variances assumption.
To get more detail on the linear model assumptions, you can refer to Idre (2017).
References
Agresti, A. (2018). Statistical Methods for the Social Sciences. 5th Edition. Pearson: Upper Saddle River, NJ, USA.
Agresti, A. and Finlay, B. (1997). Statistical Methods for the Social Sciences. 3rd ed. Pearson: Upper Saddle River, NJ, USA.
Hair, J. F., Black, W. C. Babin, B. J. and Anderson, R. E. (2014). Multivariate Data Analysis. Pearson: Upper Saddle River, NJ, USA.
Malhotra, N. K. (2009). Basic Marketing Research: A Decision Making Approach. Pearson: Upper Saddle River, NJ, USA.
Osborne and Walers (2002). Four Assumptions of Multiple Regression that Researchers Should Always Test. Practical Assesment, Research and Evaluation. 8(2), 1-5.
Idre (2017). Regression diagnostics. Retrieved from
<http://www.ats.ucla.edu/stat/spss/webbooks/reg/chapter2/spssreg2.htm> on 16 Feb 2017.
