CAEFFECT Procedure

PROC CAEFFECT Statement

PROC CAEFFECT <options>;

The PROC CAEFFECT statement invokes the procedure. Table 1 summarizes the available options in this statement. The options are described fully after the table.

Table 1: PROC CAEFFECT Statement Options

Option	Description
DATA=	Specifies the input data table
INFERENCE	Computes standard errors and confidence limits
METHOD=	Specifies the method to use for estimating potential outcome means
POMCOV	Displays the estimated covariance matrix of the potential outcome means
POMINFO	Displays a table that summarizes the potential outcome specifications
SCALEDIPWFLAG=	Specifies a value to be used to flag large weights

DATA=libref.data-table

names the input data table for PROC CAEFFECT to use. The default is the most recently created data table. libref.data-table is a two-level name, where

libref: refers to a collection of information that is defined in the LIBNAME statement and includes the library, which includes a path to the data. For more information about libref, see the section Using SAS Viya Workbench.
data-table: specifies the name of the input data table.

INFERENCE

computes standard errors and confidence intervals for the potential outcome mean and causal effect estimates. Note that because the CAEFFECT procedure takes a model-agnostic approach to causal effect estimation, the statistical validity of the inference depends on model-specific conditions that you must evaluate and that the procedure cannot verify. For more information about the conditions that are required for valid statistical inference, see the section Standard Errors and Confidence Limits.

METHOD=AIPW | IPW | REGADJ | TMLE

specifies the method to use for estimating potential outcome means (POMs). The required inputs for the procedure depend on the estimation method that you specify.

You can specify one of the following methods:

AIPW: estimates POMs by augmented inverse probability weighting. This method requires the observed values of the outcome, predicted treatment probabilities, and predicted counterfactual outcome values. For more information about the AIPW estimation method, see the section Augmented Inverse Probability Weighting.
IPW: estimates POMs by inverse probability weighting. This method requires the observed values of the outcome and predicted treatment probabilities. For more information about the IPW estimation method, see the section Inverse Probability Weighting.
REGADJ: estimates POMs by regression adjustment. This method requires predicted counterfactual outcome values. For more information about the REGADJ estimation method, see the section Regression Adjustment.
TMLE: estimates POMs by targeted maximum likelihood estimation. This method requires the observed values of the outcome, predicted treatment probabilities, and predicted counterfactual outcome values. For more information about the TMLE method, see the section Targeted Maximum Likelihood Estimation.

A default estimation method is determined by the input that you provide to the procedure. If you provide predicted treatment probabilities and predicted counterfactual outcomes, by using either an OUTCOMEMODEL statement or the PREDOUT= option in a POM statement, the TMLE method is used by default. If you provide predicted treatment probabilities and no predicted counterfactual outcomes, the IPW method is used by default. If you provide predicted counterfactual outcomes and no predicted treatment probabilities, the REGADJ method is used by default.

POMCOV

displays a covariance matrix of the potential outcome mean estimates. Note that because the CAEFFECT procedure takes a model-agnostic approach to causal effect estimation, the statistical validity of the inference depends on model-specific conditions that you must evaluate and that the procedure cannot verify. For more information about the conditions that are required for valid statistical inference, see the section Standard Errors and Confidence Limits.

POMINFO

displays the "Potential Outcome Mean Information" table. This table summarizes the levels of treatment for which potential outcome means are estimated and the names of variables that contain precomputed treatment probabilities or counterfactual outcomes that are used for estimation.

SCALEDIPWFLAG=number

specifies a multiple of the expected inverse probability weight of an observation that is used to flag observations that have large weights.

By default, SCALEDIPWFLAG=10.

Last updated: May 14, 2026