[R-sig-ME] nlmer error "pwrss failed to converge"

Martin Maechler maechler at stat.math.ethz.ch
Wed May 2 12:09:20 CEST 2018


>>>>> Louisell, Paul T PW  ........ on Wed, 2 May 2018 writes

(to the lme4-Authors)

    > Hi,
    > I'm using 64-bit R-3.5.0 on a Windows 7 platform. I'm also using the most recent version of lme4 (2018-04-03 21:41:32 UTC). Here's the error I get:

    > Error in devfun(rho$pp$theta) : prss failed to converge in 300 iterations

    > I can't find anywhere where the user has control over this parameter (prss). Is there any way to increase the number of iterations?

I assume you have looked at ?lmer which on the topic of
"control" points to ?lmerControl.
Both help pages, unfortunately, do not mention "prss" so this is
very reasonable question

However, ?lmer documents it argument `verbose` as

 verbose: integer scalar.  If '> 0' verbose output is generated during
          the optimization of the parameter estimates.  If '> 1'
          verbose output is generated during the individual PIRLS steps.

and you can take PIRLS (Penalized Iteratively Reweighted Least-Squares)
as "synonymous" here to PRSS (Penalized Residual Sum of Squares).

- In this case, you can actually not change the 300 iteration max
  (it is currently hard coded in the C++ code)

- I strongly recommend using  'verbose = 1' or even '= 2'.

- Please give us much more details, notably about your model
  formula, your data, e.g. the result of  str(.) and maybe
  summary(.)


    > Thanks in advance for any help you can give,

    > Paul Louisell
    > Statistical Specialist
    > .....

I hope the above does help further.

Also, I've taken the liberty to answer in this public
R-SIG-Mixed-Models mailing list, so others can help too and also
that others will find this answer in the future..

Best regards,

Martin Maechler
ETH Zurich



More information about the R-sig-mixed-models mailing list