No research is possible without variables. Variables exemplify one of the central components of professional qualitative and quantitative research. In this paper, a definition and description of variables is provided. A variable vs. constant dichotomy is discussed. How to decide which variables and concepts to select is discussed. The paper provides a brief description of the NOIR (nominal, ordinal, interval, ratio) levels of measurements. The concepts of validity, reliability, precision, accuracy, and their implications for professional research are discussed.
Keywords: research, variables, constants, reliability, validity, measurement.
Variables versus Constants in Research Studies
That variables are the central elements of professional research is undisputable. Much has been written and said about the importance of choosing the right variable and its implications for the reliability and validity of research findings. Internal and external threats to validity in qualitative and quantitative research have been researched in abundance. Nonetheless, not all researchers can successfully cope with the main research challenges. Understanding the meaning and significance of variables does not always suffice to develop relevant research frameworks and produce reliable, valid results. Researchers must understand how to select the most important variables, how to deal with the variables at various levels of measurement, and what to do to achieve the most acceptable level of precision, accuracy, and reliability in qualitative and quantitative studies.
Variables versus Constants: Defining and Explaining
Most, if not all, researchers think and interpret their studies in terms of variables. Whatever the type of the study, a variable is always “a construct or a characteristic that can take on different values or scores” (Ary, Jacobs, Razavieh & Sorensen, 2009, p.37). Simply put, it is a process, concept, or phenomenon which changes under the influence of various factors or causes a change. As a result, variables exemplify an interest object of professional investigation. Take, for example, drinking behaviors: they may change under the influence of numerous factors, including social status and genetic predispositions. In a similar vein, gender and intelligence can change and affect the nature of human actions, for example, employee performance in the workplace. Constants are the opposites of variables, which have fixed value (Ary et al, 2009). For example, in a study of workplace performance among marketing managers the marketing manager position/ occupation is a constant. Engel and Schutt (2005) claim that constants are not interesting to researchers. However, the current state of research lends no credibility to this assumption. On the contrary, constants are crucial to understanding the complexity of social reality. It is with the help of constants that researchers can successfully identify those variables which correspond to the concepts and setting of each study.
Variables differ by type. Independent versus dependent and categorical versus continuous dichotomies create a foundation for developing perfect research frameworks. A categorical variable is the one which represents the value of mutually exclusive categories/ groups (Ary et al, 2009). Categorical variables do not have numerical value. Gender, place of residence, language, college, and occupation are all categorical variables. Not all categorical variables have two classes, like gender. Many categorical variables cover more than one sub-category: take religious affiliation – individuals can position themselves as Christian, Jewish, Muslim, etc. In the meantime, continuous variables have an infinite number of numerical values, like height, weight, or age (Ary et al, 2009). However, this classification is not as important as the classification of variables by the way they are used in studies.
Depending on how variables are used in the study, they can be classified as independent or dependent. The former are “antecedent to dependent variables and are known or are hypothesized to influence the dependent variable, which is the outcome” (Ary et al, 2009, p.37). In medical experiments, treatment is usually an independent variable, whereas the outcomes of this treatment are dependent variables (Ary et al, 2009). In a similar vein, in a study of teacher attitudes and their effects on student achievement, teacher attitudes are antecedents of student achievement, with the latter being a dependent variable. Here, the main question is how to define the most important variables and use them appropriately. Engel and Russell (2005) offer a good solution to this problem:
- Examine and create the basic theoretical framework and define the list of theoretical concepts that are relevant to the study questions;
- Review previous knowledge and literature and estimate the utility and validity of the proposed variables;
- Identify and evaluate possible barriers to measuring these variables in the given social context; and
- Evaluate the intended role of all these variables in the study.
NOIR: Nominal, Ordinal, Interval, and Ratio Levels of Measurement
NOIR is a popular abbreviation in scholarly literature, used to describe the four levels of measurement in professional research. These levels of measurement help to define the type of variables to be used in the study and measurement procedures needed to deal with them (Smith, Gratz & Bousquet, 2008). These levels of measurement are nominal, ordinal, interval, and ratio, with each subsequent level having all properties of the level preceding it (Smith et al, 2008). The first level is the nominal level, where numbers are used to label variables: for example, female gender can be labeled as 1 and male – as 2 (Smith et al, 2008). At the next, ordinal level of measurement labels are applied to put variables in rank order (Smith et al, 2008). This level builds on and expands the nominal level of measurement and allows performing complex mathematical calculations involving ordinal scales. The ordinal level of measurement also precedes the interval level, where variables are in rank order and have equal distances between them (Smith et al, 2008). The Fahrenheit temperature is an excellent example of interval measurements, since there is always a 1-degree difference between its values. Finally, there is also the ratio level of measurement, where variables can be ranked by order, have equal distances between them, and can also have the value of an absolute zero (Smith et al, 2008). At this level of measurement, ratio comparisons finally become possible. This is the most complex level of measurement used in research.
Surprisingly or not, not all researchers accept and use this measurement typology. Velleman and Wilkinson (1993) believe that this typology is inherently misleading. The researchers are convinced that the NOIR categorization, coined by S.S. Stevens in the 1940s, fails to describe the real world and does not meet the basic criteria of modern methods of statistical analysis (Velleman & Wilkinson, 1993). Moreover, the typology creates an oversimplified picture of the statistical reality and suggests that doing statistical analysis is essentially about picking the desired level of measurement and working with it (Velleman & Wilkinson, 1993). However, in reality, these measurement scales simply shape the basis for developing effective research frameworks. Moreover, the rapid development of statistical software renders most of these claims irrelevant and invalid. Simultaneously, little doubt exists as for the importance of validity, reliability, accuracy, and precision in contemporary research. Professional researchers must have a clear understanding of these concepts, to successfully apply them in practice.
Validity, Reliability, Precision, and Accuracy
Validity is essentially about the trustworthiness and accuracy of the data, research instruments and study findings (Bernard, 2000). Validity is the most important attribute of research. Everything begins with the validity of instruments, which is integrally linked to the validity of data. With the valid instruments and data comes the validity of study findings: only conclusions reached with valid instruments and data analysis can be considered as valid. In its turn, reliability refers to whether or not researchers can obtain the same result by using the same instruments and measuring the same data more than once (Bernard, 2000). Simply stated, reliability is the same as replication, and the latter is often cited among the basic measures of research quality and professionalism. Researchers must also consider the desired degree of precision or, in other words, the number of decimal points used in measurements (Bernard, 2000). Unfortunately, even reliability, validity and precision do not guarantee accuracy. Systematic bias in professional research is not uncommon. Despite these controversies, accuracy remains central to the current study of social reality; it is the concept, the knowledge of which constantly improves (Bernard, 2000).
It is noteworthy that validity in qualitative research is somewhat different from that in quantitative studies. The meaning of validity is rooted in positivist traditions, which are integrally linked to and support the significance of other concepts, including objectivity, deduction, fact, reason, and mathematical analysis (Winter, 2000). These are the positivist traditions that gave rise to quantitative research (Winter, 2000). As a result, validity is inseparable from the need to create an objective picture of reality, with little to no personal or systematic bias. In the meantime, rooted in post-positivist epistemology, qualitative studies reject objectivity and universal truth and interpret reality through the prism of personal meanings and experiences (Winter, 2000). Therefore, validity is extremely difficult to apply in qualitative research. In the absence of universal tests and methodologies, all qualitative researchers can do is to rely on the representativeness of the sample and the quality of their findings (Winter, 2000). It would be fair to say that qualitative studies do not aim to achieve validity but, rather, help to create a unique vision of the surrounding reality. This, however, does not reduce the significance of reliability, which is equal for quantitative and qualitative studies and requires that the instruments and methodology fit in the context and meet the goals of the study.
That variables are the central elements of professional research cannot be denied. Most, if not all, researchers think and interpret their studies in terms of variables. Depending on how variables are used in the study, they can be classified as independent or dependent. Independent versus dependent and categorical versus continuous dichotomies create a foundation for developing perfect research frameworks. Yet, research is not limited to variables. Validity, reliability, precision and accuracy are all essential to achieving relevant, objective study results. Therefore, professional researchers must understand the significance of each of these concepts and the ways they must be applied in various research contexts.