The CALIS Procedure

ODS Table Name	Description	Specialized Option	Global Display Option
ContKurtosis	Contributions to kurtosis from observations	KURTOSIS	PRINT
CovRobust	Robust covariance matrix, with raw data input	PCORR	PRINT
CovDetRobust	Determinant of the robust covariance matrix, with raw data input	PCORR	PRINT
InCorr	Input correlation matrix	PCORR	PALL
InCorrDet	Determinant of the input correlation matrix	PCORR	PALL
InCov	Input covariance matrix	PCORR	PALL
InCovDet	Determinant of the input covariance matrix	PCORR	PALL
InMean	Input mean vector	PCORR	PALL
Kurtosis	Kurtosis, with raw data input	KURTOSIS	PRINT
MeanRobust	Robust mean vector, with raw data input	PCORR	PRINT
PredCorr	Predicted correlation matrix	PCORR	PALL
PredCorrDet	Determinant of the predicted correlation matrix	PCORR	PALL
PredCov	Predicted covariance matrix	PCORR	PALL
PredCovDet	Determinant of the predicted covariance matrix	PCORR	PALL
PredCovDiagnostics	Counting problematic cases of variance and covariance estimates	COVDIAG	PALL
PredMean	Predicted mean vector	PCORR	PALL
RidgedCovRobust	Ridged robust covariance matrix, with raw data input	PCORR	PRINT
RidgedCovDetRobust	Determinant of the ridged robust covariance matrix, with raw data input	PCORR	PRINT
RidgedInCorr	Ridged input correlation matrix	PCORR	PALL
RidgedInCorrDet	Determinant of the ridged input correlation matrix	PCORR	PALL
RidgedInCov	Ridged input covariance matrix	PCORR	PALL
RidgedInCovDet	Determinant of the ridged input covariance matrix	PCORR	PALL
SimpleStatistics	Simple statistics, with raw data input	SIMPLE	Default
Weights	Weight matrix	PWEIGHT	PALL
WeightsDet	Determinant of the weight matrix	PWEIGHT	PALL

1. Robust estimation with the ROBUST option is required.

ODS Table Name	Description	Specialized Option	Global Display Option
AveCoverage	Average proportion coverages of means (variances) and covariances	SIMPLE or PCORR	Default
MeanCovCoverage	Proportions of data present for means (variances) and covariances	SIMPLE or PCORR	Default
MissPatternsMeans	Means of the nonmissing and the most frequent missing patterns	SIMPLE or PCORR	Default
RankCovCoverage	Rank order of the covariance coverages	SIMPLE or PCORR	Default
RankMissPatterns	Rank order of the most frequent missing patterns	SIMPLE or PCORR	Default
RankVariableCoverage	Rank order of the proportion coverages of the variables	SIMPLE or PCORR	Default

1. You can use the NOMISSPAT option in the PROC CALIS statement to suppress the analytic output of the missing patterns. If you use the NOMISSPAT option in the GROUP statements, only the output of the missing pattern analysis for the corresponding groups are suppressed.2. PROC CALIS outputs these tables by default only when there are incomplete observations in the data sets and you use METHOD=FIML or METHOD=LSFIML for estimation.

ODS Table Name	Description	Specialized Option	Global Display Option
Residuals in the Moment Matrices
AsymStdRes	Asymptotically standardized residual matrix	RESIDUAL=ASYSTAND	PALL
AveAsymStdRes	Average of absolute asymptotically standardized residual values	RESIDUAL=ASYSTAND	PALL
AveNormRes	Average of absolute normalized residual values	RESIDUAL=NORM	PALL
AveRawRes	Average of absolute raw residual values	RESIDUAL	PALL
AveVarStdRes	Average of absolute variance standardized residual values	RESIDUAL=VARSTAND	PALL
DistAsymStdRes	Distribution of asymptotically standardized residuals	RESIDUAL=ASYSTAND	PALL
DistNormRes	Distribution of normalized residuals	RESIDUAL=NORM	PALL
DistRawRes	Distribution of raw residuals	RESIDUAL	PALL
DistVarStdRes	Distribution of variance standardized residuals	RESIDUAL=VARSTAND	PALL
NormRes	Normalized residual matrix	RESIDUAL=NORM	PALL
RawRes	Raw residual matrix	RESIDUAL	PALL
RankAsymStdRes	Rank order of asymptotically standardized residuals	RESIDUAL=ASYSTAND	PALL
RankNormRes	Rank order of normalized residuals	RESIDUAL=NORM	PALL
RankRawRes	Rank order of raw residuals	RESIDUAL	PALL
RankVarStdRes	Rank order of variance standardized residuals	RESIDUAL=VARSTAND	PALL
VarStdRes	Variance standardized residual matrix	RESIDUAL=VARSTAND	PALL
Case-Level Residuals
DepartResidualDist	Departures from the theoretical distribution for the residual M-distances	RESIDUAL	PALL
Leverage	Observations with largest leverage M-distances	RESIDUAL	PALL
Outlier	Observations with largest residual M-distances	RESIDUAL	PALL

1. The RESIDUAL= option specifies the RESIDUAL option and the type of residuals. For example, RESIDUAL=ASYSTAND requests asymptotically standardized residuals, in addition to the tables enabled by the RESIDUAL option. In some cases, the requested types of residuals are not available due to the specific estimation method or the data type used. When this occurs, PROC CALIS will determine the appropriate types of normalized or standardized residuals to display. 2. Raw residuals are also printed for correlation analysis even if RESIDUAL or PALL is not specified.3. Raw data input is required.

The ODS tables that are not model-oriented are listed in the following table:

ODS Table Name	Description	Global Display Option	Additional Specification Required
AddParms	Estimates for additional parameters	PSHORT	PARAMETERS statement
AddParmsInit	Initial values for additional parameters	PSHORT	PARAMETERS statement
Fit	Fit summary	PSUMMARY
GroupFit	Fit comparison among groups	PSUMMARY	Multiple groups
ModelingInfo	General modeling information	PSHORT
ModelSummary	Summary of models and their labels and types	PSHORT	Multiple models
ParmFunc	Parametric function testing	PSHORT	TESTFUNC statement
SimTest	Simultaneous tests of parametric functions	PSHORT	SIMTESTS statement

1. This table is displayed when you have multiple models that have labels specified by the LABEL= option, or when you define a model with more than a single level of reference by using the REFMODEL option. Otherwise, the ModelingInfo table contains all pertinent information regarding the models in the analysis.

These ODS tables are model-oriented, meaning that each model has its own set of ODS tables in the output. There are three types of model specification and results printing in PROC CALIS: initial specification, (unstandardized) estimated model results, and standardized model results. To distinguish these three types of ODS tables, different suffixes for the ODS table names are used. An "Init" suffix indicates initial specification, while a "Std" suffix indicates standardized solutions. All other tables are for unstandardized solutions.

These ODS tables require some specialized options to print. If you set the specialized option in the PROC CALIS statement, it applies to all models. If you set the specialized option in the MODEL statement, it applies to the associated model only. Alternatively, to print all these ODS tables, you can use the PSHORT or any higher level global display options in the PROC CALIS statement. Either a specialized or a global printing option is sufficient to print these ODS tables. The following is a summary of the specialized and global printing options for these three types of ODS tables:

In the following list of ODS tables, the prefixes of the ODS table names indicate the modeling language required for the ODS tables to print. The last column of the list indicates whether the PRIMAT option is needed to print the corresponding ODS tables in matrix formats. You can use the PRIMAT option either in the PROC CALIS or MODEL statement. If you want matrix output for all models, set this option in the PROC CALIS statement. If you want matrix output for a specific model, set this option in the associated MODEL statement only.

Rankings of the LM statistics in different regions of parameter space are the default printing format when you specify the MODIFICATION option in the PROC CALIS or MODEL statement. You can also turn off these default printing by the NODEFAULT option in the LMTESTS statement for models. If you want to print matrices of LM test statistics rather than the rankings of LM test statistics, you can use the NORANK or MAXRANK=0 option in the LMTESTS statement.

These ODS tables for ranking LM statistics are specific to the types of modeling languages used. This is noted in the last column of the following table.

To print matrices of LM test indices for a model, you must also use the LMMAT option in the LMTESTS statement for the model. Some of these matrices are printed by default, while others are printed only when certain regions of parameter are specified in the LM test sets. In the following tables, the ODS table names for LM test statistics in matrix form are listed for each model type.

The following table shows the ODS tables for the remaining modification indices.