Categories of Fit Indices
Absolute Fit
Measures how well the model reproduces the data directly.
Examples
Chi-square, RMSEA, SRMR
Incremental Fit
Compares your model to a null (baseline) model.
Examples
CFI, TLI (NNFI)
Parsimony Fit
Adjusts fit assessment based on model complexity.
Examples
AIC, BIC
Key Indices & Standards
| Fit Index | Good Fit Threshold | Interpretation |
|---|---|---|
| χ² (Chi-square) | p > .05 (Non-significant) | Sensitive to large samples—often significant even in good models. |
| RMSEA |
≤ .06 (good) ≤ .08 (acceptable) |
Lower is better; penalizes model complexity. |
| SRMR | ≤ .08 | Average difference between observed & predicted correlations. |
| CFI |
≥ .95 (good) ≥ .90 (acceptable) |
Compares to baseline model; values close to 1 are best. |
| TLI (NNFI) |
≥ .95 (good) ≥ .90 (acceptable) |
Similar to CFI but penalizes for model complexity. |
| AIC/BIC | Relative comparison | Lower is better. Used to compare two different models, not absolute fit. |
Visual: Interpreting RMSEA
Hover over the sections below to understand the ranges for the Root Mean Square Error of Approximation.
< .05
.05 - .08
.08 - .10
> .10
0.00
0.05
0.08
0.10
1.00
Hover over the bar above to see interpretation details.
Quick Tips
The "Golden Trio"
Always report at least three indices from different categories:
- One Absolute (e.g., RMSEA)
- One Incremental (e.g., CFI)
- One Residual (e.g., SRMR)
Conflicting Indices?
If RMSEA indicates good fit but CFI is low (or vice versa), consider:
- Sample Size
- Model Complexity
Chi-Square Warning
A significant Chi-square (p < .05) does not always mean poor fit.
In large samples (N > 200), Chi-square is almost always significant. Rely on RMSEA and CFI instead.