|
Model specification
- Use GUI or command language
- GUI uses standard path notation
- Command language natural variation on path notation
- Standard path notation generalized to allow optional access to intercepts
- Intercepts may be constrained or suppressed, just as any other variable
- Group estimation as easy as adding group(sex);
easily add or relax constraints including adding or omitting paths
for some groups but not others
Reliability
- May optionally specify fraction of variance not due to measurement
error for observed variables
Identification
- Automatic normalization (anchoring) constraints; may be overridden
- Models checked for identification
Starting values
- Automatic
- May specify for some or all parameters
- May fit one model, subset or superset, and use fitted values for
another model
Estimation methods
- ML, maximum likelihood
- MLMV, maximum likelihood for missing values; equivalent to FIML
- ADF, asymptotic distribution free, meaning GMM (generalized method of
moments) using ADF weighting matrix
- X-conditional estimation automatically used whenever appropriate, user
may override
Variance (standard error) estimation techniques
- OIM, observed information matrix
- EIM, expected information matrix
- OPG, outer product of gradients
- Robust, Huber–White sandwich estimator of variance
- Clustered, generalized Huber–White sandwich
- Bootstrap, nonparametric bootstrap
- Jackknife, delete-one jackknife
Survey support
- Sampling weights
- Stratification and poststratification
- Clustered sampling at one or more levels
Optional use of summary statistics data (SSD)
- Fit models on covariances or correlations and optionally
variances and means
- SSD may be group specific
- Easy creation and management of SSDs
- Build SSDs from original (raw) data for distribution or publication
- Automatic corruption/error checking and repairing
- Electronic signatures
Results
- May be used with postestimation features
- May be saved to disk for restoration and use later
- Optionally display results in Bentler–Weeks form
- All results accessible for user-written programs
|
Assess nonrecursive system stability
Direct and indirect effects
- Confidence intervals
- Unstandardized or standardized units
Overall goodness-of-fit statistics
- Model vs. saturated
- Baseline vs. saturated
- RMSEA, root mean squared error of approximation
- AIC, Akaike’s information criterion
- BIC, Bayesian information criterion
- CFI, comparative fit index
- TLI, Tucker–Lewis index, a.k.a. nonnormed fit index
- SRMR, standardized root mean squared residual
- CD, coefficient of determination
Equation level goodness-of-fit statistics
- R-squared
- Equation-level variance decomposition
- Bentler–Raykov squared multiple-correlation coefficient
Group level goodness-of-fit statistics
- SRMR
- CD
- Model vs. saturated chi-squared contribution
Residual analysis
- Mean residuals
- Variance and covariance residuals
- Raw, normalized, and standardized values available
Parameter tests
- Modification indices
- Wald tests
- Score tests
- Likelihood-ratio tests
- Easy to specify single or joint custom tests for omitted paths, included
paths, and relaxing constraints
- Linear and nonlinear tests of estimated parameters
- Tests may be specified in unstandardized or standardized parameter units
Group-level parameter tests
- Group invariance by parameter class or user specified
Linear and nonlinear combinations of estimated parameters
- Confidence intervals
- Unstandardized or standardized units
Predicted values
- Of observed endogenous variables
- Of latent endogenous variables
- Of latent variables (factor scores)
- Of equation-level first derivatives
- In- and out-of-sample prediction; may estimate on one sample and
form predictions in another
Explore more about SEM in Stata 12.
|
)
