topic WARNING: The covariance across equations in Statistical Procedures

WARNING: The covariance across equations

sasphd — Fri, 12 Sep 2014 18:56:50 GMT

hello,

I have this warning. How can I resolve the problem?

WARNING: The covariance across equations (the S matrix) is singular. A generalized inverse was

computed by setting to zero the part of the S matrix for the following 9 equations

whose residuals are linearly dependent with residuals from earlier equations: aone atwo

athree afour afive asix anine aten power

Re: WARNING: The covariance across equations

stat_sas — Fri, 12 Sep 2014 19:17:40 GMT

Not sure, what type of analysis is this, but most likely this is a problem of multicollinearity.

Re: WARNING: The covariance across equations

SteveDenham — Mon, 15 Sep 2014 12:22:38 GMT

It is exact collinearity, in that one of the variables is a linear combination of some the others. That is why a generalized inverse is used. You have two choices--accept the GI approach (my preference) or reparameterize your model.

Steve Denham

Re: WARNING: The covariance across equations (the S matrix) is singular...

PIRUNGU — Sun, 18 Jun 2017 11:02:26 GMT

Hi Steve,

I also got a similar message as the other member (SASPHD) above.

Can you kindly provide a syntax on how to correct for singularity using the IV approach in a source-differentiated AIDS model?

Thanks.

Patrick