AMOS Guide

Mastering AVE, CR & Fornell–Larcker

A step-by-step guide to reading AMOS output tables and transferring data into Excel for valid reliability analysis.

Start Analysis

Before You Start

Standardized Output

Analyze → Properties → Output → Check Standardized estimates before running.

Construct Order

Follow the template order exactly:
Usability (US) → Effectiveness (EF) → Engagement (EG)

Required Values

You need: Standardized loadings (λ), Error variances (θ), and Latent correlations.

Download Excel Template

Key Terms

λ
Lambda
Loading
θ
Theta
Error Variance
S.E.
Std. Error
C.R.
Critical Ratio
> 1.96
P
Significance
*** < .001

The 5 AMOS Tables (In Order)

① Regression Weights (Unstandardized)

Reference Only

Shows raw loadings with S.E. and C.R. Do not paste this into the Excel template for AVE calculation.

② Standardized Regression Weights

Action Required

Copy the Estimate column values. These are your λ (Lambda).

  • Paste to Excel sheets (US, EF, EG)
  • Column B (Loading)
Standardized Regression Weights

Check for Heywood cases (λ > 1)

③ Covariances (Latent)

Option A

Use this only if you want Excel to compute correlations for you. Paste into the Yellow lower triangle on the "Latent" sheet.

Covariances Table

④ Correlations (Latent)

Option B (Recommended)

If AMOS outputs this, paste it directly. It overwrites covariances.

Engagement ↔ Usability: .632

Engagement ↔ Effectiveness: .839

Usability ↔ Effectiveness: .695

Paste to "Latent" sheet → "Correlations (optional paste)".

Correlations Table

⑤ Variances

Action Required

Extract Item Error Variances (e1, e2...) from this table.

  • Match e-terms to items (e.g., EG1 ↔ e1).
  • Paste into Column D on respective Excel sheets.
Variances Table

Formulas & Interpretation

AVE

Convergent Validity

AVE = Σ(λ²) / [Σ(λ²) + Σ(θ)]
Target: ≥ 0.50

CR

Composite Reliability

CR = [Σ(λ)]² / {[Σ(λ)]² + Σ(θ)}
Target: ≥ 0.70

Fornell–Larcker

Discriminant Validity

√AVE(i) > |Correlation(i,j)|

The square root of AVE for a construct must be greater than its correlation with any other construct.

Quality Control & Troubleshooting

  • Values incorrect? Ensure you used Standardized Output. θ ≈ 1 − λ² is a good sanity check.
  • Item Mismatch? Check that e1 corresponds to EG1 (or your specific item). Verify in your AMOS diagram.
  • Heywood Cases: If λ > 1 or θ < 0, check for multicollinearity or model misspecification.