Mediation Analysis AMOS: How to Perform Mediation Analysis in

Mediation Analysis is one of the advanced statistical methods used in Structural Equation Modeling (SEM), and it aims to explain how the effect of the independent variable is transmitted to the dependent variable through a mediating variable. The use of AMOS software has become common in this type of analysis due to the graphical and statistical capabilities it provides, which facilitate building complex models and accurately interpreting their results.

Researchers turn to mediation analysis when the relationship between two variables is indirect, and it is assumed that one or more variables explain this relationship. This analysis is widely used in educational, psychological, administrative, and social studies, as it provides a deeper understanding of the causal relationships between variables.

In this article, we will provide a practical and methodological explanation of howto use mediation analysis in AMOSstarting from the basic concepts, through the steps of building the model and testing the indirect effect, to interpreting the results with scientific accuracy.

What Is Mediation Analysis?

Mediation analysis is a statisticalmethodused to test whether the effect of the independent variable (X) on the dependent variable (Y) occurs directly or through a mediating variable (M) that explains this relationship. In other words, this analysis aims to understand the mechanism of the effect rather than just its existence.

The mediating variable is a fundamental element in this analysis, as it clarifies the path through which the effect is transmitted, helping the researcher to provide a deeper theoretical interpretation of the relationships between variables.

The Importance of Mediation Analysis in Scientific Research

The importance of mediation analysis lies in its ability to interpret complex relationships between variables, rather than just focusing on direct relationships. It also contributes to:

1. Supporting the theoretical models proposed in the research.

2. Revealing the underlying causal mechanisms.

3. Improving the quality of statistical interpretation of results.

4. Providing more accurate results compared to traditional regression.

Components of a Mediation Analysis Model

A mediation analysis model consists of a set of essential elements, each of which forms an indispensable part in interpreting the relationship between variables within the structural model.

Independent Variable (x)

Representsthe independent variablethe influencing or causal factor that is assumed to have an effect on the dependent variable, either directly or indirectly through the mediating variable.

Mediating Variable (m)

The mediating variable serves as a link between the independent variable and the dependent variable, and is used to explain how the effect is transmitted from the independent variable to the dependent variable.

Dependent Variable (y)

The dependent variable represents the outcome or effect that the researcher seeks to explain or predict in light of the other variables in the model.

Paths in a Mediation Model

A mediation model includes four main paths:

1. Path (a): From the independent variable to the mediator variable.

2. Path (b): From the mediator variable to the dependent variable.

3. Path (c): The total effect of the independent variable on the dependent variable.

4. Path (c’): The direct effect after introducing the mediator variable.

Types of Mediation Analysis

The type of mediation analysis varies depending on the nature of the relationship between variables, and the strength of direct and indirect effects within the model.

Full Mediation

Full mediation occurs when the direct effect of the independent variable on the dependent variable becomes statistically insignificant after introducing the mediator variable, while the indirect effect is significant.

Partial Mediation

Partial mediation is achieved when the direct effect remains statistically significant, along with a significant indirect effect through the mediator variable.

Multiple Mediation

Multiple mediation is used when the model includes more than one mediator variable, allowing for the study of multiple effect paths simultaneously.

When Do We Use Mediation Analysis in AMOS?

Mediation analysis is used in AMOS when the research objective is to explain the relationship between two variables through a third variable, rather than just testing for the existence of a direct relationship. This analysis is particularly suitable in studies that are based on a theoretical framework that explains how the effect occurs.

Conditions for Using Mediation Analysis

For the use of mediation analysis to be scientifically justified, a set of conditions must be met, the most important of which are:

• ExistenceA clear theoretical frameworkthat supports the mediation hypothesis.

• Existence of a statistically significant correlation between variables.

• The possibility of theoretically and logically explaining the mediator variable.

• Data fitting for use of structural equation modeling.

The Difference Between Mediation Analysis and Traditional Regression

Although mediation can be tested using multiple regression, AMOS provides higher accuracy in estimation, as it allows testing all paths in one model, takes measurement error into account, and provides indicators of model fit quality, which is not available in traditional regression.

Assumptions of Mediation Analysis Using AMOS

Before applying mediation analysis in AMOS, it is necessary to ensure that a number of statistical and methodological assumptions are met, as violating them affects the accuracy of the results and the validity of the conclusions.

Nature of Variables

It is preferable that the variables used are quantitative and derived from measures with good psychometric properties. Latent variables can also be used, provided they have undergone confirmatory factor analysis beforehand.

Appropriate Sample Size

Mediation analysis using structural equation modeling requires a relatively appropriate sample size, where it is usually recommended that the sample size not be less than 200 cases, taking into account the complexity of the model and the number of paths.

Assumptions of Structural Equation Modeling

These assumptions include normality in distribution, no severe multicollinearity between variables, and independence of observations. Verifying these assumptions is a crucial step before building the model in AMOS.

Steps to Prepare a Mediation Analysis Model in AMOS

The model preparation stage is one of the most important stages of mediation analysis, as the quality of the results and their correct interpretation depend on it.

Preparing Data in SPSS

The process begins by examining the data inSPSSin terms of handling missing values, testing for normal distribution, and ensuring there are no outliers that affect the analysis.

Drawing the Mediation Model in AMOS

After preparing the data, the AMOS program is opened and the mediation model is drawn by creating variables and drawing paths between the independent variable, the mediator variable, and the dependent variable, according to the proposed theoretical model.

Model Specification and Running the Analysis

In this stage, the model’s characteristics are specified, such as selecting the appropriate estimation method, then the analysis is run to extract path coefficients and statistical indicators related to the model.

Testing the Indirect Effect in an AMOS Model

Testing the indirect effect is the essence of mediation analysis, as it aims to verify whether the mediator variable transmits the effect from the independent variable to the dependent variable in a statistically significant manner. AMOS is one of the best tools for testing this type of effect with high accuracy.

Using Bootstrap in Mediation Testing

AMOS relies on the Bootstrap method to test the significance of the indirect effect, as this effect does not follow a normal distribution in most cases. This method involves drawing a large number of random samples from the original data and calculating the indirect effect in each sample.

It is generally recommended to use between 2000 and 5000 Bootstrap samples for more stable and accurate results.

Interpretation of Indirect Effect Results

The significance of the indirect effect is determined through confidence intervals. If the confidence interval does not include zero, the indirect effect is statistically significant, indicating mediation.

In this context, the direct effect is not required to be significant, as mediation can occur even when the direct path is not significant.

Interpretation of Mediation Analysis Results in AMOS

After running the model and extracting the results, the next step is interpreting the statistical outputs, which requires a precise understanding of path coefficients and their significance.

Interpretation of Path Coefficients

Path coefficients represent the strength and direction of the relationship between variables in the model. Positive values indicate a direct relationship, while negative values indicate an inverse relationship. Higher coefficient values indicate stronger effects.

Significance of Direct and Indirect Effects

The type of mediation is determined based on the significance of the effects:

• If the indirect effect is significant and the direct effect is not, this indicates full mediation.

• If both effects are significant, this indicates partial mediation.

Determining the Type of Mediation

The final judgment on mediation depends on Bootstrap results and path coefficients, as well as consistency with the study’s theoretical framework. It is not advisable to rely solely on statistical significance without a supporting theoretical interpretation.

Model Fit Indices

Model fit indices are used to evaluate how well the structural model fits the data, and these indices are essential for interpreting mediation analysis results using AMOS.

Absolute Fit Indices

The most important absolute fit indices are:

• Chi-Square (χ²): It is preferred to be non-significant, though it is affected by sample size.

• RMSEA: Values below 0.08 indicate acceptable fit, and below 0.05 indicate good fit.

Comparative Fit Indices

The comparative fit indices include:

• CFI (Comparative Fit Index): Values of 0.90 or higher are preferred.

• TLI (Tucker-Lewis Index): Higher values indicate better model fit.

Acceptable Values for Fit Indices

One should not rely on a single indicator only, but it is preferable to interpret a set of indicators together, taking into account the nature of the model, its size, and the theoretical framework it is based on.

Applied Example of Mediation Analysis Using AMOS

To illustrate how to use mediation analysis in AMOS in a practical way, we will present a simplified applied example that explains the implementation steps and interpretation of results.

Description of the Research Model

Let’s assume that the researcher is studying the effect of transformational leadership (X) on job performance (Y), assuming that job satisfaction (M) plays the role of a mediating variable in this relationship. The theoretical framework assumes that transformational leadership affects job satisfaction, which in turn affects job performance.

Steps to Implement the Model

After preparing the data in SPSS, the model is drawn in AMOS through:

• Drawing the path from transformational leadership to job satisfaction.

• Drawing the path from job satisfaction to job performance.

• Drawing the direct path from transformational leadership to job performance.
  Then the Bootstrap option is activated to estimate the indirect effect and the model is run.

Displaying and Interpreting the Results

If the results show that the path from transformational leadership to job satisfaction is significant, and the path from job satisfaction to job performance is significant, and the confidence interval for the indirect effect does not include zero, this indicates that mediation is confirmed. Whether the mediation is complete or partial is determined based on the significance of the direct path.

Common Errors When Using Mediation Analysis in AMOS

Despite the relative simplicity of the procedural steps, there are a set of common errors that may affect the validity of mediation analysis results.

Ignoring the Use of Bootstrap

One of the most common errors is relying only on the significance of the paths without testing the indirect effect using Bootstrap, which may lead to inaccurate conclusions about the confirmation of mediation.

Misinterpretation of the Type of Mediation

Some researchers may make the mistake of judging complete or partial mediation without referring to the results of the indirect effect and confidence intervals, and relying only on the significance of the direct effect.

Weak Theoretical Framework

Mediation analysis should not be used without a clear theoretical basis that justifies the role of the mediating variable. Statistical significance alone is not sufficient to justify mediation from a scientific perspective.

Frequently Asked Questions About Mediation Analysis in AMOS

Can AMOS Be Used Without Latent Variables?

Yes, AMOS can be used for mediation analysis using only observed variables, but using latent variables is more accurate because it takes measurement error into account.

What Is the Difference Between Mediation and Moderation?

Mediation analysis explains how an effect is transmitted through a mediating variable, while moderation analysis studies how a moderating variable affects the strength or direction of the relationship between two variables.

Is Mediation Analysis Suitable for Master’s Theses?

Yes, mediation analysis is widely used in master’s and doctoral theses, especially in studies that rely on structural equation modeling.

What Is an Appropriate Sample Size for Mediation Analysis in AMOS?

The sample size should preferably not be less than 200 cases, taking into account the number of variables and the complexity of the structural model.

What Should I Do If the Direct Effect Is Not Significant?

The non-significance of the direct effect does not prevent mediation from occurring, as the indirect effect may be significant, which is known as complete mediation.

Conclusion of the Article

Mediation analysis using AMOS is one of the most powerful statistical methods for interpreting complex relationships between variables, as it allows the researcher to understand the mechanisms through which effects are transmitted, not just their existence. In this article, we have reviewed the concept of mediation analysis, its conditions for use, practical application steps in AMOS, as well as how to test the indirect effect and interpret the results.

The proper use of mediation analysis, supported by a clear theoretical framework and appropriate data, contributes to providing accurate and reliable scientific results, and enhances the interpretive value of academic studies, especially in the fields of humanities, social sciences, and administration.