The CALIS Procedure

You can use the AUXILIARY statement to define a set of auxiliary variables to be included in the model. The primary purpose of including auxiliary variables is to help obtain unbiased estimation when you use the full information maximum likelihood (FIML) method to analyze incomplete data. Another purpose is to make the missing at random (MAR) assumption more plausible for the use of FIML. In fact, the auxiliary variables themselves can also have missing values when you use the FIML method.

For example, the following statements specify the addition of auxiliary variables w1–w3 in the model that is specified by the PATH statement:

proc calis;
   path
      y1 <=== x1 x2,
      y2 <=== x3 x4,
      y1 <=== y2;
   auxiliary w1-w3;
run

Auxiliary variables are observed variables that are not of primary interest in your target model. They are included in the model for the sole purpose of improving estimation. PROC CALIS implements the "saturated correlates" method of Graham (2003). The auxiliary variables are treated as exogenous observed variables in the model. They generate the following additional parameters in the model:

The addition of these parameters ensures that the degrees of freedom of the original model does not change with the inclusion of the auxiliary variables.

Currently, PROC CALIS supports the use of the AUXILIARY statement only when you analyze a single sample and with METHOD=FIML. In addition, only the following types of models support the use of auxiliary variables: