Chapter 3: Research Design (to be merged with the main dissertation document)
Celine Akuns
Student ID:
Course:
University:
Table of Contents
Chapter 3: Research Design 2
3.1 Introduction 2
3.2 Research Design Modelling 3
3.2.1 Philosophy and Learning 3
3.2.2 Methodology Choice, Strategy, and Time Horizon 4
3.3 Methods 5
3.3.1 Sampling 5
3.3.2 Data Collection 6
3.3.3 Data Analysis 7
3.4 Ethical Conduct 8
3.5 Summary 8
References: 9
Chapter 3: Research Design
3.1 Introduction
A research design is an organised structure of philosophies, methodological and learning approaches, methods, materials, instrumentation, measurements, data analysis, and overall conduct (Cooper & Schindler, 2014; Saunders, Lewis, & Thornhill, 2009). The research onion model (Figure 1) presented by Saunders, Lewis, & Thornhill (2009: p. 108) and the additional concepts on research methods and ethical conduct (beyond this model) have been studied in this chapter to arrive at suitable design elements for conducting this study.
Figure 1: Research designing using the onion model by Saunders, Lewis, & Thornhill (2011: p. 108)
This chapter is presented in three sections: research design modelling, methods, and ethical conduct. These sections present the suitable design elements selected from the onion model, and the detailed steps of chose methods and ethical conduct followed in this study.
3.2 Research Design Modelling
Based on the onion model, this section presents the philosophy, learning, methodology choice, strategy, and time horizon design elements for research design modelling of this study. The reviews and choices made for these elements are presented in the Sub-sections 3.2.1 and 3.2.2.
3.2.1 Philosophy and Learning
The research philosophy determines the epistemological and ontological considerations of a research design (Bryman et al., 2011). Epistemology defines the rules of learning whereas ontology defines the structural context of the research setting in which the learning process is executed (Bryman & Bell, 2007; Bryman et al., 2011). Under epistemology, the key philosophical approaches are positivism, interpretivism, realism, and pragmatism (Saunders, Lewis, & Thornhill, 2009), and under ontology, the key philosophical approaches are objectivism and constructivism (Bryman & Bell, 2007). This sub-section evaluates epistemological philosophies only because ontology is mostly used for social and psychological research studies (Bryman & Bell, 2007).
Positivism philosophy follows the principles of natural sciences in the process of learning from any research setting (Bryman et al., 2011). The learning approach in positivism is deductive, which involves studying interrelationships between influencing and influenced variables through scientific methods for accepting or rejecting hypotheses (Bryman & Bell, 2007; Bryman et al., 2011).
Interpretivism philosophy follows the principles of experiential, observational, and evidence-based learning by applying human interpretations (hermeneutics) (Bryman et al., 2011). The learning approach in positivism is inductive, which involves exploration of knowledge through collection of subjective evidences, through observations, and by making ethnographic observations (Saunders, Lewis, & Thornhill, 2009).
Further, pragmatism follows the approach of focussing only on the research question irrespective of whether it can be answered through interpretivism (induction), positivism (deduction), or a mix of both (Saunders, Lewis, & Thornhill, 2009). Pragmatism is also open to mixing methodological approaches (qualitative and quantitative) and different methods. Just like pragmatism, realism also mixes approaches, methodologies and methods suitably (Saunders, Lewis, & Thornhill, 2009). However, realism approach assumes that a massive reality exists independent of human knowledge that needs to be explored gradually in empirical knowledge building steps.
At the outset, it appears that deciding the philosophical approach for a study is very difficult and confusing. It is very difficult to draw a hard line between two conflicting choices. However, it was certain that this research is based mostly on deductive learning approach as big data analytics, quality assurance, and supply chain management are concepts related with engineering and operations in industries. While social and human variables are also in action, most of the variables are technical. Hence, it was perceived that this research requires scientific methods thus making positivism and deduction learning the preferred choices. It is hereby emphasised that the study does not rule out interpretivism and induction completely. For example, the literature review chapter and critical discussion on concepts follow an interpretive approach with inductive learning. Perhaps, it may be wiser to state that the primary research part (data collection and analysis) shall follow positivism with deductive approach.
3.2.2 Methodology Choice, Strategy, and Time Horizon
As per the onion model, there may be three methodologies for conducting a research study: mono method (selecting either qualitative or quantitative methods), mixed method (mixing qualitative and quantitative methods as hybrid methods), and multi-method (using multiple qualitative and/or quantitative methods independently) (Saunders, Lewis, & Thornhill, 2009). In this research, a multi-method quantitative methodology was chosen following a sequence of quantitative methods presented in Section 3.3. While it is not written on stone that positivism shall only employ quantitative methods, there is a general understanding among research communities that numerical methods involving mathematics and statistics are scientific in nature suitable for deductive learning (Bryman & Bell, 2007).
The research strategy chosen is survey. Based on a detailed study of this strategy (Cooper & Schindler, 2014; Saunders, Lewis, & Thornhill, 2009), it was perceived that using structured questionnaire as an instrument for conducting a survey among quality assurance experts in supply chains will generate a uniform and consistent structure of quantitative data on the variables. Using this data, the relationships between variables in the initial theoretical construct can be studied for statistical significance using applied statistics.
Survey strategy was also chosen because of its relatively easier logistics (Cooper & Schindler, 2014; Saunders, Lewis, & Thornhill, 2009). Survey research studies have shorter time horizons as multiple respondents can answer published questionnaire (on the Internet or through e-mails) simultaneously. The time taken in answering multiple choice questions is normally much shorter than answering questions seeking detailed explanations or narrations. Hence, the primary research may be completed within a short cross section of time (like, one week), whereas fetching detailed answers (through individual or group interviews, focus group discussions, or other strategies) may require longitudinal time horizons running over multiple weeks.
3.3 Methods
The core of the research onion model represents an entire universe of data collection and analysis techniques to be chosen. The methods to be chosen are sampling, data collection, and data analysis, which are the design elements for conducting the primary research (involving the actual field work to be accomplished by the researcher) (Cooper & Schindler, 2014; Saunders, Lewis, & Thornhill, 2009). These elements are presented in the Sub-sections 3.3.1 to 3.3.3.
3.3.1 Sampling
A primary research study is conducted on a population closely influenced by the concepts under study within a research setting (Cooper & Schindler, 2014; Sekaran & Bougie, 2009). However, a research conducted on an entire population will be too expensive and cumbersome. Hence, primary research is conducted on a sample drawn from the population in such a way that the findings can be inferred back to the population with acceptable reliability and validity concerned with it.
Sampling may be conducted following non-probabilistic and probabilistic methods depending upon the choice of methodology and the sample size (Sekaran, 2003; Sekaran & Bougie, 2009). Probabilistic methods are suitable for drawing large sample sizes effectively in relatively short time periods. Non-probabilistic sampling takes much longer as decision-making on each sample member needs to be done individually.
In this research, a sample size of 100 or more was targeted to ensure better reliability and validity of statistical significances of inter-variable relationships obtained. The probabilistic method of double sampling was found as suitable. In double sampling method, a gross sample is drawn based on gross attributes of the sample and then a second sample is drawn by loading the attributes with specific details (Sekaran, 2003). The data collection method followed using double sampling is presented in the Sub-section 3.3.2.
3.3.2 Data Collection
For this research, the population comprises of individuals working for quality assurance in supply chain management (SCM) in UK and Ireland. A part of the members of this population were accessed through windows accessible to the researcher: professional and social networking websites, job portals, and professional contacts. However, quality assurance in SCM is only a draft attribute. The specific attribute needed in this research was big data-driven quality assurance in SCM. As reviewed in Chapter 2, big data is a new term coined to data analytics at large scales after fetching large volumes of data from relational and non-relation sources comprising both human- and machine-generated numbers. Hence, the specific attribute was defined as professionals experienced in heavy-duty data analytics for quality assurance in SCM.
A second-level sample comprising 146 members was drawn from the accessible sources. A link to the questionnaire was sent to the members inviting them to take part in the survey. The finalised questionnaire is presented in Appendix A. A total no. of 109 respondents accepted the invitation and answered the questionnaire. There were three incomplete responses, which were rejected in the final analysis. The final data analysis was conducted on 106 complete responses.
3.3.3 Data Analysis
The initial theoretical construct of this study derived in Chapter 2 comprised of multiple quantitative response variables as independent and dependent variables. The motivation was to study the variance of the dependent variables influenced by the variance of the independent variables. The technique of multivariable analysis of variance (MANOVA) is suitable for this kind of study (Siegel, 2012). The output of this technique not only represents statistically significant relationships between independent and dependent variables albeit also reflects a model of statistically valid relationships between a group of independent variables and another group of dependent variables (Foster, Barkus, & Yavorsky, 2006; Siegel, 2012).
In addition to model-driven multivariate analysis, there were other pre-conditions that this research had met justifying the choice of MANOVA (Foster, Barkus, & Yavorsky, 2006). First of all, the variables in the groups of independent and dependent variables were unique as they represented separate and unique concepts. Secondly, there were a large number of respondents than the number of variables. Thirdly, there were considerable variances between variables within the two groups of independent and dependent variables (that is, the variables within the groups did not have significant correlations).
The pre-step of data encoding was followed before data analysis. The data collected for each variable was encoded suitably to represent the variable (independent or dependent). Thereafter, to understand the data in its raw format, a descriptive statistical analysis was conducted reflecting its mean, standard deviation, kurtosis, and skewness (Siegel, 2012)). Finally, the probability distribution of each causal relationship between an independent and a dependent variable was studied to identify all those having the p-value equal to or less than 0.05 indicating validity of the F-statistic at 95% confidence (Field, 2009; Foster, Barkus, & Yavorsky, 2006; Siegel, 2012). The F-statistic is the key MANOVA outcome indicating the variances of dependent variable related with variances of the independent variable. F-statistic values with 95% confidence are statistically significant, which are reflected in the finalised multivariate model. All these data analytics were carried out using SPSS software for students.
3.4 Ethical Conduct
Ethics in research has a very wide scope as discussed by Saunders, Lewis, and Thornhill (2009: p. 160; 169-175). First of all, the research should be reviewed and approved by an ethical committee to analyse the possible impacts of the research. There should be no acts of discrimination, physical or psychological harm, financial harm, breach of laws and intellectual property rights, gaining material advantages, and biased reporting in the conduct of a research. Further, there should not be any unethical or illegal means to gain access to respondents, data, or materials.
This research has been conducted based on clearly outlined ethical guidelines issued by the university. There has been no form of harmful or deceptive act in this research. Only permissible sources have been studied and cited as references, and all such intellectual properties used as references have been credited using the citations and referencing guidelines issued by the university. The research supervisor played a crucial role in keeping the progress under close watch and issuing alerts and warnings as appropriate in the conduct of this research.
3.5 Summary
This chapter presents the research design elements of this study based on academically accepted empirical concepts and theories as reviewed. The philosophical approach and learning selected are positivism and deduction. Multi-method quantitative methodology was chosen with descriptive statistical analysis and MANOVA. The sampling method chosen was probabilistic double sampling and the research strategy chosen was survey. Finally, this chapter presented the ethical conduct followed in the study. The next chapter presents detailed presentation and analysis of primary data collected from survey strategy.