BookmarkSubscribeRSS Feed
brs2012
Calcite | Level 5

Hi,

I am running multivariate multilevel growth model on SAS 9.3 (x64bit) with two random components (one for intercept and one for slope). I have two dependent variables (DV). I have approximately 6000 subjects and 80000 rows or 40000 for each DV.  I used the following codes in which med is the dummy for two DVs:

proc mixed data= mult  covtest asycorr noclprint;

ods output solutionF=Fixed7 covparms=cov7

   tests3=test3_7 fitstatistics=fitstat7  gcorr=gcor7;

class enrolid med ;

model adh = med med*qtr med*age /noint s;

random med med*qtr /type=un sub=enrolid g gcorr;

repeated /type=vc group=med subject=enrolid;

run;


It computes when Random intercept model is run (i.e., if 'med*qtr' is dropped from the model and random statements). But, SAS can NOT run it as growth model and reports "integer overflow...2429.72M bytes of memory cannot be allocated" problem although I have 4 GB RAM. I tried to allot as much memory as possible but it did not work out. Please suggest any solutions or share if you see any issues.

5 REPLIES 5
SteveDenham
Jade | Level 19

If enrolid is a numeric variable, then try sorting the dataset by enrolid, and eliminating enrolid from the class statement.  If enrolid is alphanumeric, then create a nurmeric variable to use in its place.  I cannot gurantee that this will solve the problem, but it should help.

Other questions that having the answer to may help solve this: What does the variable 'qtr' represent?  Is it a successive time variable?  Since this is in proc mixed, the Kronecker product variance structures are available.  What happens (especially to the R part of the Z matrix) if you shift to:

repeated med qtr/type=un@cs subject=enrolid;

This structure is presented in the documentation for the REPEATED statement, with an example where the responses are height and weight (Var), which looks equivalent to your 'med.'

I hope this helps some.

Steve Denham

TomKari
Onyx | Level 15

I don't know if this applies to your problem, but I have seen cases with other SAS procs where they internally use signed integer for limits, so they can't assign more than 2**31 (2,147,483,648) of something. I'm not aware of any workaround for this problem.

Tom

brs2012
Calcite | Level 5

Thanks a lot guys!!

@ Steve, I took enrolid off from the class statement and replaced with a numeric one. But, it did not help.

Surprisingly, it  fails to use all of 4 GB to run when SAS default ddfm is used on the model statement. It runs successfully only when I reduce the number of subjects to 4500 (approx.) with the default containment method.

Anyway, I could finally run the model with some alternative 'ddfm= ' specifications (e.g., residual, KR, etc.) on the model statement.

Friends, thanks again for your time!!

wchang
Calcite | Level 5

Hello, your thread is the only one I found that mentions running a multivariate multilevel model. Can you direct me to any resources on how to use SAS to run this model? Thank you for your time!

Ready to join fellow brilliant minds for the SAS Hackathon?

Build your skills. Make connections. Enjoy creative freedom. Maybe change the world. Registration is now open through August 30th. Visit the SAS Hackathon homepage.

Register today!
How to Concatenate Values

Learn how use the CAT functions in SAS to join values from multiple variables into a single value.

Find more tutorials on the SAS Users YouTube channel.

Click image to register for webinarClick image to register for webinar

Classroom Training Available!

Select SAS Training centers are offering in-person courses. View upcoming courses for:

View all other training opportunities.

Discussion stats
  • 5 replies
  • 1566 views
  • 4 likes
  • 5 in conversation