Module 4: Running Linear Regression
Purpose
This module teaches how to run linear regression in SmartPLS 4 and select appropriate settings.
Learning Objectives
After this module, learners should be able to:
- Run single and multiple regression in SmartPLS.
- Choose test type and significance level.
- Choose normal, HC3, or HC4 standard errors.
- Understand standardized and unstandardized coefficients.
Regression Settings
Key settings:
| Setting | Practical Guidance |
|---|---|
| Test type | Use two-tailed unless directional hypothesis is justified before analysis |
| Significance level | Commonly 0.05 |
| Standard error type | Start with normal; use HC3/HC4 when heteroscedasticity or leverage is a concern |
| Intercept | Usually included |
SmartPLS documentation identifies HC3 and HC4 as heteroscedasticity-consistent alternatives to normal standard errors.
Unstandardized and Standardized Coefficients
Unstandardized coefficient:
One-unit change in X is associated with b-unit change in Y.
Standardized coefficient:
One standard deviation change in X is associated with beta standard deviation change in Y.
Use unstandardized coefficients for practical interpretation and standardized coefficients for relative predictor comparison.
Single vs Multiple Regression
Single regression:
Y <- X1
Multiple regression:
Y <- X1 + X2 + X3 + controls
Multiple regression coefficients are adjusted for other predictors in the model.
Practice
- Run the simple regression model.
- Run the multiple regression model.
- Export the result report.
- Compare standardized and unstandardized coefficients.
- Record R-square and adjusted R-square.
Next Module
Continue to Module 5: Assessing Regression Results.