The Core Logic
It checks whether each construct is sufficiently distinct from other constructs in the model. The logic is simple: a construct should share more variance with its own indicators than with any other construct.
// The Formula
for all \(i \neq j\)
Note: Use latent construct correlations (ϕ) from your AMOS output, not raw item correlations.
📊 Interactive Fornell–Larcker Matrix
Hover over the bold diagonal values (√AVE) to see which correlations they must exceed.
| Construct | Construct A | Construct B | Construct C |
|---|---|---|---|
| A | 0.76 (√AVE) | 0.43 | 0.51 |
| B | 0.43 | 0.72 (√AVE) | 0.39 |
| C | 0.51 | 0.39 | 0.80 (√AVE) |
Validity Checker
Enter your constructs' values below to see if they pass the Fornell-Larcker criterion.
Waiting for input...
Square Root of AVE: -
-
🧰 Workflow: From AMOS to Excel
AMOS Setup
Go to Analysis Properties → Output. Tick Standardized estimates before running the model.
Copy Data
Paste latent variances and covariances into the template's yellow cells.
Auto-Calc
The template computes AVE and CR. Optionally paste standardized loadings (λ) for precision.
The Matrix
Review the generated table. Check for the "PASS" indicator and copy to your manuscript.
Construct order: Usability → Effectiveness → Engagement
Common Pitfalls
- Low AVE (< 0.50): Weakens both convergent and discriminant validity.
- High Correlations: If |rij| ≥ √AVE, constructs overlap conceptually.
- Wrong Metrics: Avoid mixing item correlations with latent √AVE.
Quick Recap
| Metric | Purpose | Rule |
|---|---|---|
| AVE | Convergent | ≥ 0.50 |
| F-L Criterion | Discriminant | √AVE > |r| |