BookmarkSubscribeRSS Feed
SteveDenham
Jade | Level 19

Hey everyone,

 

I got tasked by the head of our neurobiology group to come up with a machine that will faithfully execute the following:

 

Group X Treatment X Day X categorical data (FOB dependent measures)

Group X Treatment X Day X continuous data (FOB dependent measures)

 

We generally have 5 distinct treatment groups, at least four treatment conditions

G1, G2, G3, G4, and G5   (16 animals per group)

 

Multiple FOBs are conducted:

 

  1. A) Day -1 [baseline},
  2. B) Day 1 [acute immediate effects [Tmax],
  3. C) 4 FOBs during chronic exposure to assess tolerance development (Day7,14,21, and 28), and then
  4. D) 7 FOBs following abrupt cessation of treatment (Day 1 withdrawal, Day 2 withdrawal, etc.)

 

Each of the FOBs have multiple dependent measures (categorical and continuous data) and, more importantly, post hoc comparisons between and within groups is critical (we now just compare against vehicle controls).

 

So, what sort of magic box to handle approximately 80 to 100 subjects, measured on 20 categorical variables and half that many continuous variables, at 13 to 15 unequally spaced time points.  We can do the univariate repeated measures, but they want to go beyond that, and take into account the correlation between the various measures as well.

 

I have written the GLIMMIX code, but it invariably dies either an "Insufficient memory" death when confronted with real data on our current machines (64 bit WIN PC w 16 GB RAM and an i7-2600CPUa 3.40 Ghz processor), or what seems like an infinite time to convergence, often with floating point overflow.

 

Steve Denham

3 REPLIES 3
PaigeMiller
Diamond | Level 26

I can't give precise answers, of course, as I've never worked on this exact problem ... however my general advice is always: get the fastest CPU you can afford and also maximum memory that the mother board can handle, and of course a 64bit operating system. Yes, I know that doesn't really help much, but that's my answer.

--
Paige Miller
Rick_SAS
SAS Super FREQ

Steve,

Sorry to hear about your massive model. Have you already read the SGF paper by Tech Support employeesTao, Kiernan, and Gibbs? http://support.sas.com/resources/papers/proceedings15/SAS1919-2015.pdf

 

There are also some related KB articles like this one: http://support.sas.com/kb/37/057.html

 

Neither of these references address your question, but rather try to recast the problem so that it requires less memory. 

data_null__
Jade | Level 19

Would you write some code to gen a data set like you describe and include the GLIMIX code you would use to analyze?  I would like to see if I can run it.

sas-innovate-2024.png

Join us for SAS Innovate April 16-19 at the Aria in Las Vegas. Bring the team and save big with our group pricing for a limited time only.

Pre-conference courses and tutorials are filling up fast and are always a sellout. Register today to reserve your seat.

 

Register now!

What is ANOVA?

ANOVA, or Analysis Of Variance, is used to compare the averages or means of two or more populations to better understand how they differ. Watch this tutorial for more.

Find more tutorials on the SAS Users YouTube channel.

Discussion stats
  • 3 replies
  • 1389 views
  • 1 like
  • 4 in conversation