Optimization Methods

PROC CALIS uses a variety of nonlinear optimization algorithms for computing parameter estimates. These algorithms are very complicated and do not always work. PROC CALIS will generally inform you when the computations fail, usually by displaying an error message about the iteration limit being exceeded. When this happens, you may be able to correct the problem simply by increasing the iteration limit (MAXITER= and MAXFUNC=). However, it is often more effective to change the optimization method (OMETHOD=) or initial values. For more details, see the section Use of Optimization Techniques on page 664 in Chapter 19, The CALIS Procedure, and refer to Bollen (1989, pp. 254 “256).

PROC CALIS may sometimes converge to a local optimum rather than the global optimum. To gain some protection against local optima, you can run the analysis several times with different initial estimates. The RANDOM= option in the PROC CALIS statement is useful for generating a variety of initial estimates.