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
- /
- SAS Programming
- /
- General Programming
- /
- Insufficient memory error in running proc glimmix

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
- Highlight
- Email to a Friend
- Report Inappropriate Content

11-25-2013 01:36 PM

Hi,

I am trying to run a repeated measures ANOVA model for a crop rotation study (split plot design) where I have (Time)18 years as repeated measure in three 6 year repeated cycles, and Input and Rotation Cycle are the treatments. The data set has 650 data points. I used the following code. I tried to use proc mixed in SAS 9.3 and it is running fine but there were normality issues even after log transformed. So I decided to use glimmix. But when tried the following model it did not complete and the " ERROR: The SAS System stopped processing this step because of insufficient memory" keeps on appearing. I tried to use a subset of data and run it but keeps on giving the same error. Also I tried to use memsize option and I am not sure how to use it. It would highly appreciated if someone can help me to get away from this.

proc glimmix data=wheat;

class Time Rep Input Rotation Cycle;

model WBM=Input Rotation Cycle Input*Cycle Rotation*Cycle Input*Rotation Input*Rotation*Cycle/ ddfm=satterth error=poisson link=log;

random Time Rep(Time) Time*Rep Time*Input Time*Rotation Time*Input*Rotation Time*Rep*Input;

random Time(Cycle)/type=UN subject=Plot(Time);

lsmeans Input|Rotation|Cycle/pdiff;

ods output diffs=ppp lsmeans=mmm;

run;

Thanks,

Dilshan

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

Posted in reply to Dilshan

11-26-2013 08:58 AM

My first thought would be to try a less specific type of covariance structure. Presumably time within cycle is equally spaced, so rather than type=UN, you may want to consider AR(1), ARH(1) or ANTE(1) as possibilites to reduce the number of parameters to be estimated. Also, since Plot is not included in the class statement, be sure that the data are sorted by time and plot. And lastly, that subject seems odd to me. If plot is nested in time, but time is nested in cycle, I think your subject might need to be plot(time cycle).

Good luck.

Steve Denham