matrix is not symmetric

[jakeberv]

Hi there,

I'm trying to learn how to use PhylogeneticEM (and l1ou). I set up a test analysis with default parameters for an scOU model as such (three traits):

phyloem.model<-PhyloEM(phylo = l1ou_test$tree, Y_data=t(l1ou_test$Y), process="scOU", parallel_alpha = TRUE, Ncores = 4)

and I'm consistently getting the following error--
@pbastide any idea how to get through it?

Thanks,
Jacob Berv

Error in { : task 9 failed - "matrix is not symmetric [1,2]"
In addition: Warning message:
In find_grid_alpha(phylo, alpha, nbr_alpha = nbr_alpha, allow_negative = allow_negative, :
The chosen alpha_max was above the machine precision. Taking alpha_max as the largest possible on this machine.

[pbastide]

Hi Jacob,

Thank you for considering PhylogeneticEM for your analysis, and sorry you are experiencing problems !

The error is not very specific (which is my fault). In the past, this kind of error has been linked with the structure of the tree or the data (e.g. two traits highly correlated almost to one). Do you see anything in the data that might look "odd" ?

It can also be linked with the version of the package you use. Do you have the last CRAN version of PhylogeneticEM (1.5.0) ? Also which version of the Matrix package do you have ?

Sorry I cannot be much more help without a more detailed description. If would be great if you could isolate a "minimal working example" reproducing the error. If you can send me such an example (here or by email), I'd be happy to look into it.

Sorry for the inconvenience, and thank you again for your feed back !

Best,
Paul

[jakeberv]

Hi Paul — Thanks for getting back! I am not sure there are obvious weird aspects of the data. The tree I’m using does have some short internal branch lengths, but not sure if that would be problematic. I have tried with two or three traits (for example, body mass and generation length — which are well correlated but not perfectly correlated (ie, close to one as you suggest). The procedure also seems to fail with less obviously correlated traits, however the example dataset for the scOU model does seem to work— so this would point toward some aspect of the input data (to what I am not sure). Here are the package versions I have loaded — it looks like these are the most recent versions. I am running R 4.1.1 on MacOS Mojave.

packageVersion("Matrix")

[1] ‘1.3.4’

packageVersion("PhylogeneticEM")

[1] ‘1.5.0.9000’ Do you have any other obvious troubleshooting steps to try? Otherwise I am happy to send an example. Is your email address indicated here the appropraite one to use? http://pbastide.github.io/ <http://pbastide.github.io/> Jake

…

On Nov 15, 2021, at 3:27 AM, Paul Bastide ***@***.***> wrote: Hi Jacob, Thank you for considering PhylogeneticEM for your analysis, and sorry you are experiencing problems ! The error is not very specific (which is my fault). In the past, this kind of error has been linked with the structure of the tree or the data (e.g. two traits highly correlated almost to one). Do you see anything in the data that might look "odd" ? It can also be linked with the version of the package you use. Do you have the last CRAN version of PhylogeneticEM (1.5.0) ? Also which version of the Matrix package do you have ? Sorry I cannot be much more help without a more detailed description. If would be great if you could isolate a "minimal working example" reproducing the error. If you can send me such an example (here or by email), I'd be happy to look into it. Sorry for the inconvenience, and thank you again for your feed back ! Best, Paul — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#20 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AKMB23VF6QCUVPFSVI5SS4TUMC76VANCNFSM5H3K4KUQ>. Triage notifications on the go with GitHub Mobile for iOS <https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675> or Android <https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub>.

[pbastide]

Hi Jake,

Thank you for the feedback !

Sorry, I'm running out of simple explanations, I'm not sure what could be causing the error.

One thing you could try is to change the tree and use the same data (e.g. take a random tree with the same species, or even a star tree), or use the same tree but different data (e.g. simulated) to see if one of the two components is causing the error... But I'm not sure that it would give good results.

Otherwise, if you have an small example that fails fast, I'm happy to look into it. The address you mention should work.

Sorry for the trouble, and thank you for trying to make it work !

Paul