"Not enough memory for Kenward and Roger" error during linear mixed effects modeling

methomas — Wed, 22 Nov 2023 21:35:42 GMT

I am running a linear mixed effects model with 28,966 rows of data, where I am nesting the random effects with 4 levels: CombineMatch[CellType,MouseName,FieldOfView], comparing across three groups (CellType), and I have a covariate (Distance). I have not been able to run the analysis due to the error: “Not enough memory for Kenward and Roger. Try the Variance Components Only option.”

I have been able to successfully run this model on smaller datasets, so I think it is the size of the data and the complexity of the CombineMatch variable in particular that is causing problems. CombineMatch represents the measures that have been repeated across experiments, so there can be between 1-5 rows corresponding to each CombineMatch value.

I have tried to run with the box "Estimate Only Variance Components" checked, but the output does not contain the information I need to evaluate my model. When I try to run using the EMS Method, the data processes overnight and finally JMP becomes unresponsive and I need to force quit.

I am using JMP version 16.2.0 on Windows 10.

Does anyone have advice for how to process this dataset?

Re: "Not enough memory for Kenward and Roger" error during linear mixed effects modeling

txnelson — Mon, 27 Nov 2023 12:49:56 GMT

Honestly, memory is very inexpensive these days. I would look into adding memory to your computer if possible.

topic Re: "Not enough memory for Kenward and Roger" error during linear mixed effects modeling in Discussions

"Not enough memory for Kenward and Roger" error during linear mixed effects modeling

Re: "Not enough memory for Kenward and Roger" error during linear mixed effects modeling