Proc Causalmed

[violet77]

Hello,

I am using proc causalmed for survey data and would like to include sampling weights. However, I am unsure where to include the weight statement. I tried to include it after the covar statement but receive an error in SAS saying that the weight statement was out of order. Thanks!

PROC CAUSALMED DATA=;
CLASS;
MEDIATOR;
MODEL;
COVAR;

[sbxkoenk]

Hello,

I have never used PROC CAUSALMED (although I know what it is meant for) but I do not think the WEIGHT statement should get a particular place in the order of statements. Same for the COVAR statement.

I think it's best that you show us the LOG.

When you reply, hit the </> (Insert Code) button in the toolbar and paste the LOG in the pop-up window.

That way the LOG preserves structure and formatting which makes it easier to read.

Thanks,

Koen

[FreelanceReinh · Accepted solution]

Hello @violet77,

The WEIGHT statement of PROC CAUSALMED was introduced in SAS/STAT 15.1. I'm using SAS/STAT 14.3 as part of SAS 9.4M5, where it was not available yet (compare the 14.3 documentation to the 15.1 documentation). So, for obvious reasons, I get the error message

20      weight w;
        ------
        180
ERROR 180-322: Statement is not valid or it is used out of proper order.

which I suspect is the same as you got.

Please check your SAS version with

%put &sysvlong;

or the SAS/STAT version with

proc product_status;
run;

I'm afraid I wouldn't know a workaround (other than an upgrade) if your version was the same as mine or older.

[violet77]

Thanks! I have SAS/STAT 14.3. I appreciate your help! I will see if I can upgrade to 15.1

[violet77]

Hello,

I was able to get SAS/STAT 15.1. However, I ran into another issue. When I estimated my model with the weight statement, I received this message:

WARNING: The covariance matrix for the outcome model is not positive definite. The convergence is questionable.
WARNING: The covariance matrix for the mediator model is not positive definite. The convergence is questionable.
WARNING: Standard errors are not computed for the effect estimates.

Thanks!

[FreelanceReinh]

This is most likely an issue with your input data. Please open a new thread (e.g., with subject line "Proc Causalmed: covariance matrix not positive definite") in this forum ("Statistical Procedures") so that all experts will see it. I'm still on SAS/STAT 14.3 and hence couldn't run your code even if I had similar input data.

Good luck!