NEW AND IMPROVED NORMAL THEORY (EM-TYPE MISSING DATA) METHODS

• Jamshidian-Bentler EM-type missing data procedures for one or multiple samples.

• Kim-Bentler test of missing completely at random (MCAR), including homogeneity of means and covariances.

• LM and Wald tests in multi-sample analysis.

• Bentler-Yuan test for potential structured mean models.

• Bentler-Raykov corrected R-square for nonrecursive models.

• Regression and Bentler-Yuan optimal GLS factor scores computed and saved.

• Advanced start values for structured mean analysis.

• Standard errors for total effects.

• Internal consistency reliability and maximal reliability coefficients for composite based on a 1-factor model.

• Reliability coefficient rho for a factor model.

• Cronbach’s alpha, greatest lower bound reliability, Bentler’s dimension-free lower bound reliability and Shapiro’s lower bound reliability for a weighted composite.

CORRECTED NORMAL THEORY METHODS FOR NON-NORMAL & MISSING DATA

• Satorra-Bentler statistic for multiple methods and multi-sample analysis.

• Correct (including Satorra-Bentler robust) standard errors for indirect and total effects.

• Satorra-Bentler information matrix for LM test.

• Yuan-Bentler F-test and Yuan-Bentler-Browne residual-based statistics.

• Yuan-Bentler correct statistics for non-normal missing data.

HETEROGENEOUS KURTOSIS METHODS

• Bentler, Berkane and Kano statistics for heterogeneous kurtoses.

ARBITRARY DISTRIBUTION METHODS

• Asymptotically distribution free (ADF) mean and covariance structure methodology.

• Yuan-Bentler corrected chi-square and F-tests.

• Yuan-Bentler corrected ADF standard errors.

• ADF analysis of correlation structures.

CASE (SUBJECT) WEIGHTING METHODS

• A priori case weights for weighted mean and covariance structure analysis (e.g., for complex sample surveys).

• Yuan-Bentler case-robust methodology for outliers and influential observations.

MULTI-LEVEL MODELS

• HLM-like (Chou, Bentler and Pentz) multilevel  methodology for latent variables.

• Bentler-Liang maximum likelihood multilevel methodology.

• Muthén’s MUML multilevel methodology.

RESAMPLING AND SIMULATION METHODS AND STATISTICS

• Standard errors for standardized solution etc via bootstrap.

• Model based bootstrapping (extended Beran-Bollen-Stine methodology).

• Expanded simulation capacities: multiple group and MCAR data generation.

• Deng-Lin FMRG random number generator.

MODELING FEATURES AND APPROACHES

•  /MODEL paragraph for simple model specification (virtually eliminates the need for /EQUATION, /VARIANCE, and /COVARIANCE sections).  Many equations are built with a few simple script commands.

• Command script for building many constraints simply.

• Maximum number of model variables (including categorical) raised to 200.

• SAVE paragraph for saving imputed data and factor scores.

• Statistics and indexes for Satorra-Bentler "robust" statistics reportable separately.

• EQS output optional in HTML format.

• EQS output optional in matrix format or compact format (instead of equation format).

DIAGRAMMER

• Advanced Diagrammer for creating and reporting a model.

• Wizard system to create path, factor, structural equation, and latent growth curve models.

• Polynomial-orthogonal coefficients computed for latent growth curve model.

USER INTERFACE IMPROVEMENTS

• Project manager for organizing and retrieving all analyses from a treeview outline.

• Transporter for moving a project and its related analyses with a simple drag and drop operation.

• More usable data editor with unlimited sample size.

• Data sheet modifiable with a simple drag and drop operation.

• Covariance matrix is now entered directly on the data editor.

• Better 3D data plot.

• Improved and more general ANOVA.

• Some non-parametric general statistics analyses available.

• Front-end EM missing data procedure for data imputation.

• Improved equation builder.

• General-purpose multi-equation data transformation in a syntax window.  

• Transformation formulas are savable and re-usable.  

• Simple creation of z-scores for a variable or an entire data sheet.