Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.

Showing results for

Find a Community

- Home
- /
- SAS Programming
- /
- SAS Procedures
- /
- Problem using Proc GLIMMIX

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

07-28-2011 08:52 AM

Hi,

When I was trying to fit Proc Glimmix model, I found the following message repeatedly.

"**WARNING: Obtaining minimum variance quadratic unbiased estimates as starting values for the**

** covariance parameters failed**."

My SAS code is the following.

ods html;

ods graphics on

proc glimmix data=malaria INITGLM NOBOUND asycov asycorr oddsratio ORDER=DATA MAXOPT=100 METHOD=RSPL;

class sex s1_03 s1_12 s1_07 s2_01 s2_02 s2_03 s2_04 s2_05 s2_06 rm1_08 rm1_09 rm1_10 rm1_14 n01_115t;

model rdt_res = age sex famsize s1_03 s1_13 s2_01 s2_02 s2_03 s2_04 s2_05 s2_06 tot_room rm1_08

rm1_09 rm1_10 rm1_11 rm1_12 rm1_14 rm1_nets n01_115t s2_01*rm1_09 s2_02*rm1_10

age*sex sex*s2_01 sex*s2_04 sex*rm1_10 /dist = binary solution covb corrb DDFM=BETWITHIN;

random intercept/ subject = s1_07;

Covtest GLM ; Covtest DIAGG ; Covtest DIAGR ; Covtest HOMOGENEITY ; Covtest START ;Covtest ZEROG ;

weight Weight;

run;

ods graphics off;

ods html close;

I need help. Please could any one give me a solution?

Regards,

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

07-28-2011 10:21 AM

This is far from a solution. I count 15 classification variables (I don't know how many levels for each, and some of which do not appear in the model statement, e.g. s1_12), and 8 continuous covariates (age, famsize, s1_13, tot_rm, rm1_11, rm1_12, rm1_nets, and age*sex). Some of these look like classification variables as well, but you may well wish to treat them as continuous.

So, I wonder about two things--complete collinearity (or separation) between some of the continuous covariates, and whether there is enough data to estimate everything.

Given that these are not problems, what happens if you delete the INITGLM option? You may still get mivque(0) problems, but it is a cheap try. I would also try dropping the ddfm=betwithin, as I don't see any within subject variability being modeled. As a last resort, try running this in PROC MIXED and save the covariance parameters to a data set that you call as starting values with a PARMS statement.