COVAR effects;
The COVAR statement specifies the effects of covariates in a causal mediation analysis. These covariates represent important confounders in the causal model. You do not need to distinguish confounders for the treatment-outcome, treatment-mediator, or mediator-outcome relationships. You simply enter all confounding covariate effects in this statement. PROC CAUSALMED appends these effects into the design matrices of the outcome and mediator models that you specify in the MODEL and MEDIATOR statements, respectively. The causal mediation and other related effects are thus estimated with adjustment for confounding by including covariate effects in outcome and mediator modeling.
The simplest form of effects is a list of confounding covariates. For example, the following statement specifies that C1, C2, and C3 are confounding covariates in the causal model:
covar C1 C2 C3;
You can also include interaction terms in the specification. For example, the following statement adds the interaction of C1 and C2 as a confounding effect to the preceding specification:
covar C1 C2 C3 C1*C2;
Alternatively, you can use the following equivalent specification:
covar C1|C2 C3;
If a confounding covariate represents nominal (classification) data, you must also include the covariate in the CLASS statement. For more information about specifying effects, see the section Specification of Effects in Chapter 53, The GLM Procedure.