To better understand the syntax of PROC CALIS, it is useful to classify the statements into classes. These classes of statements are described in the following sections.
The PROC CALIS statement is the main statement that invokes the CALIS procedure. You can specify options for input and output data sets, printing, statistical analysis, and computations in this statement. The options specified in the PROC CALIS statement will propagate to all groups and models, but are superseded by the options specified in the individual GROUP or MODEL statements.
The GROUP statement signifies the beginning of a group specification. A group in the CALIS procedure is an independent sample of observations. You can specify options for input and output data sets, printing, and statistical computations in this statement. Some of these group options in the GROUP statement can also be specified in the MODEL or PROC CALIS statement, but the options specified in the GROUP statement supersede those specified in the MODEL or PROC CALIS statement for the group designated in the GROUP statement. For group options that are available in both of the GROUP and PROC CALIS statements, see the section Options Available in the GROUP and PROC CALIS Statements. For group options that are available in the GROUP, MODEL, and PROC CALIS statements, see the section Options Available in GROUP, MODEL, and PROC CALIS Statements. If no GROUP statement is used, a single-group analysis is assumed. The group options for a single-group analysis are specified in the PROC CALIS statement.
The GROUP statement can be followed by subsidiary group specification statements, which specify further data processing procedures for the group designated in the GROUP statement.
Subsidiary group specification statements are for specifying additional data processing attributes for the input data. These statements are summarized in the following table:
| Statement | Description |
|---|---|
| AUXILIARY | Specifies the auxiliary variables |
| FREQ | Specifies the frequency variable for the input observations |
| PARTIAL | Specifies the partial variables |
| VAR | Specifies the set of variables in analysis |
| WEIGHT | Specifies the weight variable for the input observations |
These statements can be used after the PROC CALIS statement or each GROUP statement. Again, the specifications within the scope of the GROUP statement supersede those specified after the PROC CALIS statement for the group designated in the GROUP statement.
The MODEL statement signifies the beginning of a model specification. In the MODEL statement, you can specify the fitted groups, input and output data sets for model specification or estimates, printing options, statistical analysis, and computational options. Some of the options in the MODEL statement can also be specified in the PROC CALIS statement. These options are called model options. Model options specified in the MODEL statement supersede those specified in the PROC CALIS statement. For model options that are available in both of the MODEL and PROC CALIS statements, see the section Options Available in the MODEL and PROC CALIS Statements. If no MODEL statement is used, a single model is assumed and the model options are specified in the PROC CALIS statement.
Some of the options in the MODEL statement can also be specified in the GROUP statement. These options are called group options. The group options in the MODEL statement are transferred to the groups being fitted, but they are superseded by the group options specified in the associated GROUP statement. For group options that are available in the GROUP and the MODEL statements, see the section Options Available in GROUP, MODEL, and PROC CALIS Statements.
The MODEL statement itself does not define the model being fitted to the data; the main and subsidiary model specification statements that follow immediately after the MODEL statement do. These statements are described in the next two sections.
Main model specification statements are for specifying the type of the modeling language and the main features of the model. These statements are summarized in the following table:
| Statement | Description |
|---|---|
| COSAN | Specifies general mean and covariance structures in matrix terms |
| FACTOR | Specifies confirmatory or exploratory factor models |
| LINEQS | Specifies models by using linear equations |
| LISMOD | Specifies models in terms of LISREL-like model matrices |
| MSTRUCT | Specifies parameters directly in the mean and covariance matrices |
| PATH | Specifies models by using the causal paths of variables |
| RAM | Specifies models by using RAM-like lists of parameters |
| REFMODEL | Specifies a base model from which the target model is modified |
You can use one of these statements for specifying one model. Each statement in the list represents a particular type of modeling language. After the main model specification statement, you might need to add subsidiary model specification statements, as described in the following section, to complete the model specification.
Subsidiary model specification statements are used to supplement the model specification. They are specific to the types of the modeling languages invoked by the main model specification statements, as shown in the following table:
| Statement | Specification | Modeling Languages |
|---|---|---|
| COV | Covariance parameters | FACTOR, LINEQS |
| MATRIX | Parameters in matrices | COSAN, LISMOD, MSTRUCT |
| MEAN | Mean or intercept parameters | FACTOR, LINEQS, PATH |
| PCOV | (Partial) covariance parameters | PATH |
| PVAR | (Partial) variance parameters | FACTOR, PATH |
| RENAMEPARM | New parameters by renaming | REFMODEL |
| VARIANCE | Variance parameters | LINEQS |
Notice that the RAM modeling language does not have any subsidiary model specification statements, because all model specification can be done in the RAM statement.
Model analysis statements are used to request specific statistical analysis, as shown in the following table:
| Statement | Analysis |
|---|---|
| DETERM | Sets variable groups for computing the determination coefficients; same as the STRUCTEQ statement |
| EFFPART | Displays and partitions the effects in the model |
| FITINDEX | Controls the fit summary output |
| LMTESTS | Defines the Lagrange multiplier test regions |
| SIMTESTS | Defines simultaneous parametric function tests |
| STRUCTEQ | Sets variable groups for computing the determination coefficients; same as the DETERM statement |
| TESTFUNC | Tests individual parametric functions |
Notice that the DETERM and the STRUCTEQ statements function exactly the same way.
Optimization statements are used to define additional parameters and parameter constraints, to fine-tune the optimization techniques, or to set the printing options in optimization, as shown in the following table:
| Statement | Description |
|---|---|
| BOUNDS | Defines the bounds of parameters |
| LINCON | Defines the linear constraints of parameters |
| NLINCON | Defines the nonlinear constraints of parameters |
| NLOPTIONS | Sets the optimization techniques and printing options |
Other statements that are not listed in preceding sections are summarized in the following table:
| Statement | Description |
|---|---|
| BY | Fits a model to different groups separately |
| OUTFILES | Controls multiple output data sets |
| PARAMETERS | Defines additional parameters or superparameters |
| SAS programming statements | Define parameters or functions |
Note that SAS programming statements include the ARRAY statement and the mathematical statements for defining parameter interdependence.