singular convergence error in multilevel model

[Bonnie Dixon]

Hi all.

I am getting a singular convergence error in a multilevel model I am trying to run using the package, nlme. Can anyone explain what a singular convergence is and what I can do to avoid this error? Here is what I did:

require(nlme)

control.list <- 
  lmeControl(maxIter = 500, msMaxIter = 500, msMaxEval=500,
             tolerance = 0.1, msTol = 0.1)

ps.ls.gender.cb.age.wpbh <- 
  lme(efficiency ~ 
        prior.sleep + sleepiness + gender + circ.bedtime + age.18 + WP.hunger, 
      ~ prior.sleep + sleepiness + circ.bedtime + WP.hunger | person, 
      data = mod.dat, na.action = na.exclude, method = "ML", 
      control = control.list)

And this is the error message I got:

Error in lme.formula(efficiency ~ prior.sleep + sleepiness + gender +  : 
  nlminb problem, convergence error code = 1
  message = singular convergence (7)

I was trying to use the lmeControl() settings to avoid the problem, but I have to admit that I don’t really know what I am doing with that. I also tried running the model with the default method, REML, but got the same error.

Let me know if anyone has any ideas. Thanks!

Bonnie

​

[Michael Hannon]

Hi, Bonnie. This article:

https://stat.ethz.ch/pipermail/r-help/2006-January/086368.html

states that:

In general, "singular convergence" means you were trying to estimate
k parameters when the data would support estimation k-1 or fewer.

-- Mike

> --
> Check out our R resources at
> http://www.noamross.net/davis-r-users-group.html
> ---
> You received this message because you are subscribed to the Google Groups
> "Davis R Users' Group" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to davis-rug+...@googlegroups.com.
> Visit this group at http://groups.google.com/group/davis-rug.
> For more options, visit https://groups.google.com/d/optout.

[Bonnie Dixon]

Thanks for the insight, Mike.  I finally figured out how to get my model to work by including the argument, sing.tol=1e-20, in lmeControl().

Bonnie