TUTORIAL SEM PLS: METODE ANALISIS DATA

Summary

This video tutorial explains the Structural Equation Modeling (SEM) method for data analysis, building on the previously discussed multiple linear regression method. SEM allows researchers to analyze complex relationships between variables more efficiently. The presenter highlights key concepts like outer and inner models, validity and reliability tests, and demonstrates using Smart PLS software to conduct SEM analysis. The tutorial covers steps for managing models, testing hypotheses, and generating regression equations. The video also outlines future tutorials focusing on SEM with intervening and moderating variables. The tutorial aims to simplify SEM for research purposes.

Takeaways

• 📊 Structural Equation Modeling (SEM) is a multivariate analysis technique used for examining complex relationships between variables, both direct and indirect.
• 🔄 SEM allows researchers to analyze relationships in one test, unlike multiple regression, where relationships are tested separately.
• 📉 SEM is especially helpful when dealing with multiple variables, including intervening or moderating variables.
• 📈 SEM includes testing for outer models and inner models, where outer models focus on indicator validity, and inner models test reliability and structural relationships.
• 📝 Convergent validity is evaluated by outer loading values, where a value greater than 0.5 is considered valid.
• 🔍 Discriminant validity ensures that indicators for one latent variable are distinct from others, using AVE (Average Variance Extracted) values greater than 0.5.
• ⚠️ Collinearity is checked using the VIF (Variance Inflation Factor), where values greater than 5 indicate a potential issue.
• 🛠 Reliability is tested through Cronbach's Alpha and Composite Reliability, with acceptable values being above 0.7.
• 🔬 R-squared values show how well the independent variables predict the dependent variables, with higher values indicating better predictive power.
• 📊 Hypothesis testing in SEM uses P-values and T-statistics, where P-values less than 0.05 indicate a significant relationship between variables.

Q & A

What is Structural Equation Modeling (SEM)?

Structural Equation Modeling (SEM) is a multivariate analysis technique that allows researchers to test complex relationships between variables, including both unidirectional and bidirectional relationships, to obtain a comprehensive understanding of a model. According to Ghozali (2008), SEM can simultaneously test both structural models and measurement models.

How does SEM differ from multiple linear regression?

SEM allows for the simultaneous analysis of multiple relationships between variables, including both direct and indirect effects, while multiple linear regression analyzes one relationship at a time. SEM can handle more complex models with intervening and moderating variables, making it more flexible compared to the more rigid approach of multiple linear regression.

What are the steps involved in SEM analysis using PLS (Partial Least Squares)?

The steps in SEM using PLS include: 1) Testing the outer model, which focuses on the reliability and validity of the measurement model, and 2) Testing the inner model or structural model, which examines relationships between latent variables. Key tests include convergent validity, discriminant validity, and collinearity.

What is the purpose of convergent validity in SEM?

Convergent validity is used to demonstrate that each indicator can effectively explain its latent variable. An outer loading value of 0.5 or higher is considered strong evidence of convergent validity, according to Hair et al. (2010).

How is discriminant validity assessed in SEM?

Discriminant validity tests whether indicators of a specific latent variable differ from those of other latent variables. It can be evaluated using the Average Variance Extracted (AVE) values, where the AVE should be greater than 0.5. Additionally, the square root of the AVE should be higher than the correlation between the latent variable and other variables.

What does a Variance Inflation Factor (VIF) indicate in SEM?

The VIF measures collinearity between indicators within a latent variable. A VIF value below 5 indicates that there is no problematic collinearity between indicators, ensuring that the model's predictive power remains stable and reliable.

How do you assess reliability in SEM models?

Reliability is assessed using Cronbach's Alpha and Composite Reliability, where values should be greater than or equal to 0.7. Indicators with these values are considered reliable, and in the PLS software, this is indicated when the values appear in green.

What does the R-Squared value represent in SEM?

The R-Squared value represents how well the independent variables explain the variance in the dependent variable. A higher R-Squared value indicates that the model has stronger predictive power. For example, an R-Squared value of 0.513 means that 51.3% of the variance in the dependent variable is explained by the independent variables.

How is hypothesis testing conducted in SEM?

Hypothesis testing in SEM involves checking the P-value and the T-statistic. A P-value below 0.05 indicates that the relationship is statistically significant, while the T-statistic is compared against a critical value (such as 1.661 for a 95% confidence level) to further confirm significance.

What happens if an indicator in SEM has an outer loading below 0.5?

If an indicator has an outer loading below 0.5, it is considered weak in terms of validity and can be removed from the model. The model is then recalculated to ensure all remaining indicators have an acceptable level of outer loading (above 0.5).