Comparison of the TCALIS and SYSLIN Procedures

The SYSLIN procedure in SAS/ETS software can also fit certain kinds of path models and linear structural equation models. PROC TCALIS differs from PROC SYSLIN in that PROC TCALIS is more general in the use of latent variables in the models. Latent variables are unobserved, hypothetical variables, as distinct from manifest variables, which are the observed data. PROC SYSLIN allows at most one latent variable, the error term, in each equation. PROC TCALIS allows several latent variables to appear in an equation—in fact, all the variables in an equation can be latent as long as there are other equations that relate the latent variables to manifest variables.

Both the TCALIS and SYSLIN procedures enable you to specify a model as a system of linear equations. When there are several equations, a given variable might be a dependent variable in one equation and an independent variable in other equations. Therefore, additional terminology is needed to describe unambiguously the roles of variables in the system. Variables with values that are determined jointly and simultaneously by the system of equations are called endogenous variables. Variables with values that are determined outside the system—that is, in a manner separate from the process described by the system of equations—are called exogenous variables. The purpose of the system of equations is to explain the variation of each endogenous variable in terms of exogenous variables or other endogenous variables or both. Refer to Loehlin (1987, p. 4) for further discussion of endogenous and exogenous variables. In the econometric literature, error and disturbance terms are usually distinguished from exogenous variables, but in systems with more than one latent variable in an equation, the distinction is not always clear.

In PROC SYSLIN, endogenous variables are identified by the ENDOGENOUS statement. In PROC TCALIS, endogenous variables are identified by the procedure automatically after you specify the model. With different modeling languages, the identification of endogenous variables by PROC TCALIS is done by different sets of rules. For example, when you specify structural equations by using the LINEQS modeling language in PROC TCALIS, endogenous variables are assumed to be those that appear on the left-hand sides of the equations; a given variable can appear on the left-hand side of at most one equation. When you specify your model by using the PATH modeling language in PROC TCALIS, endogenous variables are those variables pointed to by arrows at least once in the path specifications.

PROC SYSLIN provides many methods of estimation, some of which are applicable only in special cases. For example, ordinary least squares estimates are suitable in certain kinds of systems but might be statistically biased and inconsistent in other kinds. PROC TCALIS provides three major methods of estimation that can be used with most models. Both the TCALIS and SYSLIN procedures can do maximum likelihood estimation, which PROC TCALIS calls ML and PROC SYSLIN calls FIML. PROC SYSLIN can be much faster than PROC TCALIS in those special cases for which it provides computationally efficient estimation methods. However, PROC TCALIS has a variety of sophisticated algorithms for maximum likelihood estimation that might be much faster than FIML in PROC SYSLIN.

PROC TCALIS can impose a wider variety of constraints on the parameters, including nonlinear constraints, than can PROC SYSLIN. For example, PROC TCALIS can constrain error variances or covariances to equal specified constants, or it can constrain two error variances to have a specified ratio.