The heteroscedasticity-consistent standard errors computed by PROC REG from the ACOV and HCC options in the MODEL statement are incorrect if the model is not full rank. The following notes are displayed in the output when the model is not full rank:

     NOTE: Model is not full rank. Least-squares solutions for the parameters are not unique. 
           Some statistics will be misleading. A reported DF of 0 or B means that the 
           estimate is biased.

     NOTE: The following parameters have been set to 0, since the variables are a 
           linear combination of other variables as shown.

A linear regression model is not full rank, or singular, when an effect in the model can be expressed as a linear combination of other effects in the model. This redundancy can happen inadvertently or by including extra dummy variables in the model. You might see an incorrect standard error estimate of 0 for one or more of the variables involved in the singularity. The standard errors for the other variables in the model are correct.

For a discussion of the less than full rank parameterization that occurs with dummy variable coding, see Usage Note 22585.

To circumvent the problem, omit the redundant effect(s).

Click the Hot Fix tab in this note to access the hot fix for this issue.

Operating System and Release Information

A fix for this issue for SAS/STAT 12.3 is available at:

A fix for this issue for SAS/STAT 12.1 is available at: