Cholmod 'out of memory' error in mixed model linear regression

steffen_munch · Post by **steffen_munch** » Sat Dec 16, 2023 11:58 am

Hi,
I am running a mixed model linear regression in gamlj 2.6.6 with 5 factors and a cluster variable with around 100 different expressions.
When I let it run on my Computer (16GB RAM I get an error in the means of Cholmod error 'out of memory' at file ../Core/cholmod_memory.c, line 147. When I use another Computer with just 8GB RAM it says cannot allocate vector of 9.8 GB. I havent tried more RAM yet, because I wanted to post this here first and I'm not sure this issue would be fixed by just shooting more Power at it.
My data is around 36.000 lines. When I use just a fraction (5000 lines), I get a working mixed model and results without any error.
The file is too big to be attached here, bit I could send it via fileshare if needed.

Any help would be very much appreciated

Steffen

mcfanda@gmail.com · Post by **mcfanda@gmail.com** » Sun Dec 17, 2023 12:22 pm

36000 rows shouldn't be a problem. Maybe it is the model that is too complex. How does your model look like?

steffen_munch · Post by **steffen_munch** » Mon Dec 18, 2023 8:40 am

Hi, thanks for the reply.
I honestly couldn't tell. It does not seem very complex to me though. Code is the following. With around 100 expressions in the cluster variable 'Messnummer'

The code in syntax mode is this:

gamlj::gamljMixed(
formula = AR ~ 1 + Geschlecht + Behandlung + Sauerstoffentzug + Messzeitpunkt + Altersgruppen + Sauerstoffentzug:Messzeitpunkt+( 0+Messzeitpunkt | Messnummer ),
data = data,
re_listing = "main",
qq = TRUE)

mcfanda@gmail.com · Post by **mcfanda@gmail.com** » Mon Dec 18, 2023 11:59 am

would you share the data so I can check?

jamovi