Nested communityPGLMM error in random effects

[kspeeriful]

Hi!

I'm trying to run a PGLMM using 5 random effects, one of which is nested. However, I keep receiving the below error:

Error in chol.default(re.i[[3]]) :
the leading minor of order 3 is not positive definite

After sequentially removing the other random effects, the single nested random effect is the one causing the error (it is the 3rd in the list of 5, which is what I assume re.i[[3]] indicates). I looked through stack overflow and found that this error can mean that the matrix you provided is not square. The random effect is below:

r.intercept.bacphy_bf<-list(1, sp=dat$Genus, covar=Vphy, site=dat$Sample)

There is no missing data and I checked that sp and site are the same length, so I don't see why they asymmetrical. Is there some other meaning to this error? Or am I missing something about the input format required for nested random effects?

Thanks for any insight!
Kelly

[arives]

Kelly, Square matrices can be non-positive definite. When a matrix is not positive definite, it means that it is not a covariance matrix. For example, the matrix 1 -.2 -.2 -.2 1 -.2 -.2 -.2 1 is not positive definite, because it is impossible to have three variables that all have pairwise negative covariances. This is just a complicated way of saying that for some reason your nested covariance matrix isn't formulated correctly. Diajiang Li and I have been working on an update of communityPGLMM in the package phyr. To load it, use: library(devtools) devtools::install_github("daijiang/phyr") You can find a lot of examples in chapter 4 of a book you can download for free: https://leanpub.com/correlateddata This might be the easiest way of side-stepping the problem. Cheers, Tony

…

______________ Anthony R. Ives UW-Madison 459 Birge Hall 608-262-1519 From: Kelly Speer <notifications@github.com> Reply-To: willpearse/pez <reply@reply.github.com> Date: Monday, March 25, 2019 at 2:30 PM To: willpearse/pez <pez@noreply.github.com> Cc: Subscribed <subscribed@noreply.github.com> Subject: [willpearse/pez] Nested communityPGLMM error in random effects (#70) Hi! I'm trying to run a PGLMM using 5 random effects, one of which is nested. However, I keep receiving the below error: Error in chol.default(re.i[[3]]) : the leading minor of order 3 is not positive definite After sequentially removing the other random effects, the single nested random effect is the one causing the error (it is the 3rd in the list of 5, which is what I assume re.i[[3]] indicates). I looked through stack overflow and found that this error can mean that the matrix you provided is not square. The random effect is below: r.intercept.bacphy_bf<-list(1, sp=dat$Genus, covar=Vphy, site=dat$Sample) There is no missing data and I checked that sp and site are the same length, so I don't see why they asymmetrical. Is there some other meaning to this error? Or am I missing something about the input format required for nested random effects? Thanks for any insight! Kelly — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub<#70>, or mute the thread<https://github.com/notifications/unsubscribe-auth/ALF_LL5Q7AFFn9MkhII8fInp7P5hej4Rks5vaSPagaJpZM4cIoYR>.

[kspeeriful]

Hi Dr. Ives,

Thank you so much for the helpful guidance and for recommending the book chapter! This book seems the exact type of guide that I was looking for! The data I'm analyzing are the microbiome compositions of ectoparasites of bats found in different habitat fragments of varying sizes, so a multi-tier hierarchy to put it nicely. I look forward to reading your book and developing my understanding of the statistical underpinnings of PGLMM!

Thanks again,
Kelly

[arives]

Kelly, Sounds like a great project. Let me know how it goes – and what I could do to improve the methods and book. Cheers, Tony From: Kelly Speer <notifications@github.com> Reply-To: willpearse/pez <reply@reply.github.com> Date: Monday, March 25, 2019 at 10:04 PM To: willpearse/pez <pez@noreply.github.com> Cc: "Anthony R. Ives" <arives@wisc.edu>, Comment <comment@noreply.github.com> Subject: Re: [willpearse/pez] Nested communityPGLMM error in random effects (#70) Hi Dr. Ives, Thank you so much for the helpful guidance and for recommending the book chapter! This book seems the exact type of guide that I was looking for! The data I'm analyzing are the microbiome compositions of ectoparasites of bats found in different habitat fragments of varying sizes, so a multi-tier hierarchy to put it nicely. I look forward to reading your book and developing my understanding of the statistical underpinnings of PGLMM! Thanks again, Kelly — You are receiving this because you commented. Reply to this email directly, view it on GitHub<#70 (comment)>, or mute the thread<https://github.com/notifications/unsubscribe-auth/ALF_LBlco2r9waIMpGFjELGWCzUnMUWMks5vaY44gaJpZM4cIoYR>.