Measures of Model Fit Flashcards
1
Q
NFI
A
- Bentler-Bonnet Normed Fit Index
- compares X2 of model to X2 of independence model
- > .95 is good, but may underestimate fit in small samples
- comparative
2
Q
CFI
A
- Bentler’s Comparative Fit Index
- similar to NFI but adjusts for non-centrality
- > .95 is good
- comparative
3
Q
RMSEA
A
- Root mean square approximation
- estimates lack of fit compared to perfect fit (saturated model)
- comparative
4
Q
GFI
A
- goodness of fit
- analogous to R2 (0 bad, 1 good)
- variance explained fit measures
5
Q
AGFI
A
- adjusts GFI for no. of parameters in the model
- improved fit with more parameters
- variance explained fit measures
6
Q
PGFI
A
- parsimonious adjustment to CFI
- parsimonious
7
Q
AIC
A
- Aikake Information Criteria
- not an absolute value, just a value of improvement between models
- want decreases between models
- parsimonious
8
Q
RMR
A
- root mean square residual
- average difference b/w sample and estimated covariance matrices
- small = good fit
- residual based fit indices
9
Q
SRMR
A
- standardised RMR
- scale 0-1
- residual based fit indices
10
Q
CAIC
A
- consistent aikake information criteria
- parsimonious
11
Q
Comparative
A
- compare model to baseline
- NFI
- CFI
- RMSEA
12
Q
Absoulte
A
broken into goodness-of-fit and residual based
13
Q
Goodness-of-fit
A
- proportion of covarainces in sample data explained by model
- GFI
- AGFI
14
Q
Residual-based
A
- difference b/w observed and predicted covariances
- RMR
- SRMR
15
Q
Parsimony
A
- corrections made to “punish” more complex models
- PGFI
- AIC
- CAIC
- RMSEA
