turn on suggestions

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

Showing results for

Find a Community

- Home
- /
- Analytics
- /
- Stat Procs
- /
- Re: GLIMMIX does cannot calculate minimum variance...

Topic Options

- 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
- Permalink
- Email to a Friend
- Report Inappropriate Content

3 weeks ago

Hi all,

I am trying to build a hierarchical model for a large survey dataset and I keep getting the error "WARNING: Obtaining minimum variance quadratic unbiased estimates as starting values for the covariance parameters failed" when fitting the following model:

```
proc glimmix data = arrlib.data pconv = 1e-4;
class hospid flag_chf(ref = '0');
model flag_ihca(event = '1') = flag_chf / cl dist = binary oddsratio solution;
random intercept / sub = hospid type = chol;
nloptions maxiter = 100 gconv = 1e-4;
weight discwt;
by flag_cancer;
title 'small data';
run;
```

However, when I fit the model using a subset of this data, I am able to get the OR with no problems. I'm not sure if this is arising from the number of observations and whether I need to allow more iterations, or what.

Thanks for y'all's consideration.

Accepted Solutions

Highlighted
## Re: GLIMMIX does cannot calculate minimum variance quadratic unbiased estimates

Options

Solution

2 weeks ago

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

Posted in reply to benbuck

2 weeks ago

Hi all,

I noticed this post gathered attention but no replies. I tweaked my modeling further and solved the problem. In case it helps anyone else, I've inserted my updated code which works:

```
proc glimmix data = arrlib.all_hosp_cancer noclprint;
class hospid predictor(ref = '0');
model flag_ihca(descending event = '1') = / ddfm = bw dist = binary solution oddsratio;
random intercept / sub=hospid ;
title1 'One-way ANCOVA with random effects for Mortality';
title2 'Building basic model'; nloptions maxiter = 10000 gconv = 1e-4 tech = nrridg ;
ods output covparms = covparmestimate;
run;
proc glimmix data = arrlib.all_hosp_cancer noclprint;
class hospid predictor(ref = '0');
model flag_ihca(descending event = '1') = predictor / ddfm = bw dist = binary solution oddsratio;
random intercept / sub=hospid ;
title1 "One-way ANCOVA with random effects for Mortality";
title2 "Testing &predictor. in &cancer.";
nloptions maxiter = 100000 maxfunc = 200000 gconv = 1e-4 tech = nrridg ;
parms / pdata = covparmestimate;
ods output OddsRatios = ORs_predictor;
run;
```

All Replies

Highlighted
## Re: GLIMMIX does cannot calculate minimum variance quadratic unbiased estimates

Options

Solution

2 weeks ago

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

Posted in reply to benbuck

2 weeks ago

Hi all,

I noticed this post gathered attention but no replies. I tweaked my modeling further and solved the problem. In case it helps anyone else, I've inserted my updated code which works:

```
proc glimmix data = arrlib.all_hosp_cancer noclprint;
class hospid predictor(ref = '0');
model flag_ihca(descending event = '1') = / ddfm = bw dist = binary solution oddsratio;
random intercept / sub=hospid ;
title1 'One-way ANCOVA with random effects for Mortality';
title2 'Building basic model'; nloptions maxiter = 10000 gconv = 1e-4 tech = nrridg ;
ods output covparms = covparmestimate;
run;
proc glimmix data = arrlib.all_hosp_cancer noclprint;
class hospid predictor(ref = '0');
model flag_ihca(descending event = '1') = predictor / ddfm = bw dist = binary solution oddsratio;
random intercept / sub=hospid ;
title1 "One-way ANCOVA with random effects for Mortality";
title2 "Testing &predictor. in &cancer.";
nloptions maxiter = 100000 maxfunc = 200000 gconv = 1e-4 tech = nrridg ;
parms / pdata = covparmestimate;
ods output OddsRatios = ORs_predictor;
run;
```