topic Re: PROC SURVEYREG - How to deal with a note in the log regarding singular matrix in Statistical Procedures

PROC SURVEYREG - How to deal with a note in the log regarding singular matrix

dode — Wed, 08 Nov 2017 01:16:16 GMT

Hello

i am using proc surveyreq because my data is survey data (complex survey design) i am using (health and retirement study data).

my hypothesis is there is association between medication regimen complexity and cognitive status in older adults

The code is below but the problem is when i run the code, i can see a note in the result section that seems that something is wrong,

proc surveyreg data=transform;
class a4dl i4adl gender m4s yearsofedu i4ncome race h7ealth c7over h4osstay c4ondition d4ep l4iving;

model logc4omp= R7COGTOT R7AGEY_B gender m4s yearsofedu i4ncome race h7ealth c7over h4osstay c4ondition d4ep l4iving a4dl i4adl / solution adjrsq;/*putt all of the here cont and cata*/

cluster raehsamp;

strata raestrat;

weight r7wtresp;
output out=surveyreg r=r p=p;
/*proc gplot data= surveyreg; plot r*p;*/

run;

this is what i see, do i ignore it or i need to fix something ?

Note:

The denominator degrees of freedom for the t tests is 52.
Matrix X'WX is singular and a generalized inverse was used to solve the normal equations. Estimates are not unique."

how I fix the problem ?

Re: PROC SURVEYREG - How to deal with a note in the log regarding singular matrix

Reeza — Wed, 08 Nov 2017 01:17:30 GMT

Note: I've edited your post to clarify your subject and format your code. I've also moved it into the Statistical Procedures forum (like before). In general, it's not a great idea to post duplicate questions.

Re: PROC SURVEYREG - How to deal with a note in the log regarding singular matrix

SAS_Rob — Wed, 08 Nov 2017 17:05:36 GMT

You can basically ignore the message and. It is a result of using the CLASS statement and the necessity of creating a series of dummy variables. SURVEYREG over-parameterizes the model such that it sets the last level to zero, essentially making it a reference category, and then estimates the other parameters as compared to the referent.

It is more thoroughly explained in this usage note

http://support.sas.com/kb/22585

Although that refers specifically to Proc GLM, what is said holds for SURVEYREG as well.

Re: PROC SURVEYREG - How to deal with a note in the log regarding singular matrix

dode — Mon, 13 Nov 2017 13:07:47 GMT

So I don't need to recode my catagorical variables? Or remove the class statement? What about it's says in the note that "estimates is not unique , do I still ignore it ?

Re: PROC SURVEYREG - How to deal with a note in the log regarding singular matrix

SAS_Rob — Mon, 13 Nov 2017 13:12:11 GMT

No, you do not need to recode or change anything. It is just a note letting you know that if you chose to use different dummy variables you would end up with a different set of estimates, even though the fit would be the same.

Re: PROC SURVEYREG - How to deal with a note in the log regarding singular matrix

dode — Mon, 13 Nov 2017 13:15:49 GMT

So I consider the beta estimates and the p value, I can take them as my results ... right ?

Re: PROC SURVEYREG - How to deal with a note in the log regarding singular matrix

SAS_Rob — Mon, 13 Nov 2017 14:29:11 GMT

Yes