Error in %*%: non-conformable arguments

[Dimitris Poursanidis]

Just run a complex maxnet model with 0.5 beta increase from 1 to 5.

In finish and then i have this error at the console >

Warning: Error in %*%: non-conformable arguments

  95: predict.maxnet

  94: maxnet::maxnet

  93: modelTune.maxnet

  92: tuning

  91: ENMeval::ENMevaluate

  90: <reactive> [modules/mod_c6_maxent.R#94]

  74: mod.maxent

  73: observeEventHandler [C:\Users\dpour\Documents\R\win-library\3.5\wallace\shiny/server.R#519]

   2: shiny::runApp

   1: run_wallace

Any idea on?

[Gonzalo Pinilla-Buitrago]

Hi Dimitris,

This issue looks like an ENMeval error instead of a Wallace error. Do you try with the maxent.jar option?

Let's wait that some of the developers of ENMeval package (some of them are in this google group) can give you some light on this problem.

Best,

Gonzalo

[Jamie M. Kass]

Sorry this is so late, but was this resolved? If not, could you provide more details on your Wallace run and/or make a reproducible example in code? Thanks.

Jamie

[Cory Merow]

Hi all,

I'm 99% sure this is a maxnet bug. A colleague and I noted it and didn't come up with a real solution. It occurs in maxnet when it computes mm %*% object$betas where object is the model and betas are the coefficients. The cause is that there are sometimes duplicate names in model$betas, and thus duplicate terms in the formula. And when the model matrix (mm) is computed using model.matrix() it seems to remove, or ignore, the duplicate term, so the dimensions of the model matrix, specifically the number of columns, is not equal to the length of object$betas, and the error occurs. I *think* there must be a a stochastic component to the optimization algorithm in maxnet because I use a catch (in R code, not Wallace) to note this issue and just try rerunning the model til it doesn't throw an error. And that seems to work. If you find that rerunning the model a few times reproduces the issue, I'd be interested to know.

Cory

Cory Merow

Ecology and Evolutionary Biology

University of Connecticut

Homepage

Google Scholar

appear.in/cmerow

skype: cory.merow

'Give me 4 parameters and I will draw you an elephant. Give me 5 and I will wag the trunk.'

--
You received this message because you are subscribed to the Google Groups "Wallace Ecological Modeling App" group.
To unsubscribe from this group and stop receiving emails from it, send an email to wallaceEcoMo...@googlegroups.com.
To post to this group, send email to wallac...@googlegroups.com.
Visit this group at https://groups.google.com/group/wallaceEcoMod.
To view this discussion on the web visit https://groups.google.com/d/msgid/wallaceEcoMod/b591142a-8685-4351-9e4b-9c95287aa4b6%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Jamie M. Kass]

Weird... thanks for the info, Cory. Would this easily be solved by a pull request for an error catch?

Jamie

[Dimitris Poursanidis]

and it happens again!!

three times in the row with maxnet...

shall i switch to maxent.jar to complete the modelling work or try again? 

[Dimitris Poursanidis]

forgot to paste the R console msg

*** Running ENMeval v2.0.4 with maxnet from maxnet package v0.1.4 *** |=========== | 12%Warning: Error in %*%: non-conformable arguments 2: shiny::runApp 1: run_wallace

[Jamie M. Kass]

One way to avoid this I've found is to avoid H1 parameterization because it often occurs when there are many possible features. However, for the long term, it would be good to post this as an issue to the maxnet repo, ideally with Cory's old explanation of what seems to be wrong. Maybe Steven can fix it.

Jamie

To view this discussion on the web visit https://groups.google.com/d/msgid/wallaceEcoMod/b7436b78-edd4-4d32-94a6-c3f26fda9888n%40googlegroups.com.

[Dimitris Poursanidis]

Thanks, I'll look into it. 

---

".. If something has to be done, take the responsibility and do it.."

Interpreting Earth & Ocean and its elements using satellites, models & field data

Dimitris Poursanidis, Ph.D. Marine Ecology