Model Comparison
VerifiedFactor Analysis
Run this test straight away on a free built-in teaching dataset — no data of your own needed — or bring your own. Either opens the guided workspace: variable setup, assumption diagnostics, results with effect sizes and confidence intervals, figures, and APA-ready reporting.
Loading teaching datasets…
Or use your own dataset
Loading your datasets…
Compares competing models on fit versus parsimony to pick the best-supported one.
Reports the fit-index battery (χ², CFI, TLI, RMSEA, SRMR) and information criteria (AIC, BIC) to judge absolute fit and compare nested or non-nested models — better fit and lower AIC/BIC win.
Worked example
Does a one-factor or two-factor model fit better?
One- and two-factor CFA models were compared on fit indices and information criteria (n = 400).
The two-factor model fit far better than the one-factor model (CFI = .99 vs .47) and was retained.
Model comparison strongly favoured the two-factor model (CFI .99 vs .47).
Try it yourself: Load this ready-made sample and follow the run above.
When to use it
- Choosing among competing modelsTwo or more specifications of the same data — you want the fit-index battery and information criteria side by side. e.g. one-factor vs two-factor CFA.
- Balancing fit against parsimonyAIC and BIC reward fit but penalise extra parameters, so a slightly worse-fitting simpler model can win.
When NOT to — use instead
- Only one model to judgeFor absolute fit of a single model, read its own fit indices rather than a comparison. → Confirmatory factor analysis
- Comparing measurement across groupsTo test whether a model holds across groups, use invariance testing. → Measurement invariance
Assumptions (and what to do if they fail)
Check: AIC/BIC are only comparable when the models are fit to the identical cases and variables (same N, no listwise-deletion differences).
If violated: A model that quietly dropped fewer cases looks better for the wrong reason — the criteria are not comparable.
Check: Nested models can use a χ² difference test; non-nested ones rely on AIC/BIC and fit indices, not a significance test.
If violated: A χ² difference test applied to non-nested models is invalid.
Check: Report the difference (ΔAIC/ΔBIC); a tiny gap is weak evidence even though one number is lower.
If violated: Declaring a winner on a trivial ΔAIC over-reads noise as a decision.
Ready to run a Model Comparison on your own data?
Guided setup, automatic assumption checks, effect sizes, figures and an APA write-up.
Run this test →