Module 5: Assessing Regression Results

Purpose

This module teaches how to assess SmartPLS regression output systematically.

Learning Objectives

After this module, learners should be able to:

• Interpret coefficient direction, size, and significance.
• Assess R-square and adjusted R-square.
• Inspect diagnostic plots.
• Identify multicollinearity and unstable estimates.

Regression Assessment Order

1. Confirm the model ran successfully.
2. Check sample size after missing-data handling.
3. Check descriptive statistics.
4. Assess collinearity among predictors.
5. Interpret unstandardized coefficients.
6. Interpret standardized coefficients.
7. Assess p-values, t-values, and confidence intervals.
8. Assess R-square and adjusted R-square.
9. Inspect QQ plot and residual diagnostics.
10. Decide whether robust standard errors are needed.

Coefficient Interpretation

Template:

Holding the other predictors constant, a one-unit increase in X was associated with a b-unit change in Y.

Example:

Holding ease of use, trust, service quality, age, and experience constant, perceived usefulness was positively associated with satisfaction.

R-Square

R-square represents the proportion of variance in the dependent variable explained by the predictors.

Adjusted R-square penalizes unnecessary predictors and is usually better for comparing models with different numbers of predictors.

Diagnostics

Check:

• QQ plot for residual normality
• Residual patterns
• Influential cases if available
• Multicollinearity among predictors
• Heteroscedasticity concerns

Red Flags

• Significant coefficient with wrong theoretical sign
• Very high standard errors
• Predictors highly correlated with each other
• R-square very high but weak theory
• Results change strongly after adding controls
• Reporting p-values without effect interpretation

Lab

Complete Lab 3: Assess regression output.

Next Module

Continue to Module 6: Regression Bootstrapping.