The increasing accessibility to data and its potential in social survey research is the key motivation for presenting a summary of the variables usually used in quantitative research. The main objective of this article is to present a guide to assist students doing research at the Bachelors and Master level with an interest in quantitative research to understand the key variables required in their research. The first section describes the method and data analysis used in the survey research. The types of variables are presented in section 2. The difference between mediation and moderating variable is presented in the third section, and the conclusion is presented in the fourth section.
Methods and Data Analysis Used in Survey Research
The key factor in determining the number of variables in research studies depends on the research question. The research question determines the number of variables required in the study. A research question with one variable (univariate) would use frequency distribution as a method of data analysis. A bivariate relationship consists of two variables. For example, if the research question is finding the relationship between servant leadership and project success, the bivariate method would be used since there are two variables. If three or more variables are present in the research question, the multivariate method is appropriate approach. Table 1 presents the number of variables and the related data analysis techniques.
|Univariate||Bivariate Methods||Multivariate Methods|
|Frequency distributions||Cross tabulations|
Correlation matrix Regression
Rank order correlation
Comparison of means
|Multiple and partial regression|
moderate regression analysis,
Modelling Relationships among the Variables
The relationship between variables is of the key interest to students conducting quantitative research. The current article focuses on variables are that are recommended at the Bachelor and Master levels. The variables are: (i) antecedent variable; (ii) dependent variable; (iii) moderating variables and (iv) mediating variable. Table 2 presents definition of the variables.
|Variable||Independent variable causes a change in the dependent variable and also known as the predictor variable.|
|Antecedent||The cause of an independent variable|
|Independent (IV)||Independent variable causes a change in the dependent variable and also known as the predictor variable.|
|Dependent variable (DV)||The dependent variable is the effect, and depends on the changes of the independent variable. Dependent variables are also known as the outcome variable.|
|Moderator||A moderator variable influences the relationship between an independent and a dependent variable. A moderator causes the relationship between a dependent/independent variable to change, depending on the value of the moderator variable.|
|Mediating variable||A mediator is the reason for the effect and acts like a “go-between” in the relationship between independent and dependent variable.|
Identifying Moderating and Mediating Variables
The key differences between moderators and mediators are presented in Figure 1. An antecedent variable is a variable that affects the independent variable.
Figure 1. Modelling relationship between variables.
A moderator is a variable that causes the relationship between two variables to change when it is systematically changed (Stone, 1978). Shields and Shields (1998) describe a moderator as a variable that affects the relationship between an independent and a dependent variable, it is not a cause of a dependent variable as is an independent variable, but it is theorized to affect the relationship between an independent and a dependent variable. Sharma et al. 1981 proposes a framework for determining the presence of moderator variable and identified two types of moderators. A pure moderator variable is defined as having nonsignificant, bivariate relationships with both the independent and dependent variables. A quasi moderator is one which the moderator interacts significantly with the independent variable and the dependent variable. For example, Agbejule and Saarikoski (2006) investigated the relationship between budget participation (BP), cost management knowledge and managerial performance. In their cost management is the moderating variable between the relationship between budget participation and managerial performance. The moderation model shows that the more cost management knowledge a manager possess, the better the participating in budget has impact on performance.
Figure 2. Illustration of moderating model (Agbejule and Saarikoski, 2006).
Baron and Kenny (1986) describe the following conditions for observing mediation:
- the independent variable (X) significantly affects the mediator (M);
- X significantly affects the dependent variable (Y) in the absence of M;
- M has a significant unique effect on Y; and
- the effect of X on Y decreases upon the addition of M to the model
An illustration of mediating variable is presented in Figure 2 below. Agbejule et al (2021) show that organizational climate mediates the relationship between trust and team learning. Following the definition of Baron and Kenny (1986), trust (independent variable, X) is significantly correlated to organizational climate (mediator variable, M). Trust significantly affects team learning (dependent variable), and the effect of trust on team learning decreases when the organizational climate is added to the model.
The objective of the article was to present a simplified approach for students to understand the basic modelling of relationship between variables when conducting a quantitative study. The paper summarizes the different data analysis techniques to be employed, depending on number variables to be analysed in the study. This would enable students to focus on the appropriate statistical method when analysing the data of the research. With the abundant and growing importance of big data, students should be encouraged to develop their statistical skills, since it can provide opportunities for thesis work in data analysis. Both methods can be used with SPSS; however, mediation models are easier with AMOS statistical package.