Skip to content

R-sig-mixed-models Digest, Vol 85, Issue 6

6 messages · Cole, Tim, David Duffy, Emmanuel Curis +1 more

Original
David Duffy
#
On Wed, 8 Jan 2014, Cole, Tim wrote:

            

      
      
      
    

        
In your example, it looks like quasi-separation for mat v. DecAge in males 
might be to blame (when I plot them), and:

library(glmmML)
glmmML(mat ~ log(DecAge) + Sex * Ethnicity, cluster=BHID, data=df,
        control = list(maxit = 2000))
[glmmml] fail = 1
Max. No. of iterations reached without convergence

Cheers, David Duffy.

| David Duffy (MBBS PhD)
| email: David.Duffy at qimrberghofer.edu.au  ph: INT+61+7+3362-0217 fax: -0101
| Genetic Epidemiology, QIMR Berghofer Institute of Medical Research
| 300 Herston Rd, Brisbane, Queensland 4006, Australia  GPG 4D0B994A

  
  


  


      

    

    

      
      

        


  

        

      Emmanuel Curis
    

    

      
      #
    

  


  

      
Hi,

I'm not sure if this may be related or not to the issue, but in the
subset you sent you seem to have duplicate rows, differing only by
their row name (see for instance 3090 & 3697, 3156 & 3763). I guess
this is because some columns were removed, but could it be related to
the difficulties to fit, and the oddities of error occuring or not
according to the subset ?

Best regards,

        
On Tue, Jan 07, 2014 at 02:12:13PM +0000, Cole, Tim wrote:

        
? Thanks Ben for your speedy response, and for confirming that my
problem is new.

? The dataset has 5048 records and 607 subjects. I've experimented with subsets of the data and it is rather complicated - I'll send you the data and code separately. The error message is simple to generate with a trivial dataset, restricted to subjects with 2 to 4 records and both 0 and 1 values for mat. But there is also a problem with subjects that have 10 records, though not 9 or 11 ? very odd.
? 
? mat is a binary mature bone measure, which for each subject consists of zero or more 0s followed by zero or more 1s, the 0/1 transition providing useful information about the age of maturity.
? 
?      BHID  Sex Ethnicity    DecAge mat
? 2     102 Girl     White  9.273973   0
? 4251  102 Girl     White 15.520000   1
? 55    169 Girl     White  9.805381   0
? 1876  169 Girl     White 12.876781   0
? 3090  169 Girl     White 14.846575   1
? 3697  169 Girl     White 14.846575   1
? 676   184 Girl     White 10.425736   0
? 6139  184 Girl     White 19.452430   1
? 1896  192  Boy     White 12.493616   0
? 6145  192  Boy     White 19.430527   1
? 1307  356  Boy     Black 11.246575   0
? 1914  356  Boy     Black 12.307274   0
? 6163  356  Boy     Black 19.255305   1
? 3156  386  Boy     Black 14.890000   0
? 3763  386  Boy     Black 14.890000   0
? 4977  386  Boy     Black 17.130000   0
? 6191  386  Boy     Black 19.364819   1
? 156   421 Girl     Black  9.235616   0
? 6226  421 Girl     Black 19.526352   1
? 264   531 Girl     Black  9.799984   0
? 1478  531 Girl     Black 11.796624   0
? 2692  531 Girl     Black 13.763384   0
? 5120  531 Girl     Black 16.752000   1
? 1103  778 Girl     White 10.732375   0
? 1710  778 Girl     White 11.747646   0
? 6566  778 Girl     White 19.433265   1
? 1791  869 Girl     Black 11.968995   0
? 3005  869 Girl     Black 13.997055   0
? 4826  869 Girl     Black 16.073000   0
? 6040  869 Girl     Black 18.124000   1
? 4834  880  Boy     White 15.504000   0
? 5441  880  Boy     White 17.076000   1
? 6048  880  Boy     White 17.823000   1
? 6655  880  Boy     White 18.562628   1
? 
? Best wishes,
? Tim
?

        
? On 14-01-06 04:04 AM, Cole, Tim wrote:

        
? The dataset consists of longitudinal measures of bone score during
? puberty, where 1000 indicates maturity. The aim is to estimate median
? age at maturity in four groups by sex and ethnicity.
? This glm code works fine, but ignores the longitudinal element. lm2
? <- glm(I(RUSBoneScore == 1000) ~ log(DecAge) + Sex * Ethnicity,
? family=binomial, data=na.omit(bh[, 3:6]))
? This code, which adds a random subject effect, fails with the
? unhelpful error message below. lm3 <- glmer(I(RUSBoneScore == 1000) ~
? log(DecAge) + Sex * Ethnicity + (1 | BHID), family=binomial,
? data=na.omit(bh[, 2:6]))
? Error in pwrssUpdate(pp, resp, tolPwrss, GHrule(0L), compDev,
? verbose) : c++ exception (unknown reason)
? I've tried various alternatives but they all fail in the same way.
? Thoughts please.
? 
?     Hard to say without a reproducible example. My main comment is that
? this is *not* a problem I have seen anyone report before.  sessionInfo()
? please?  Can you try with verbose=100? Can you make a reproducible
? example that is a reasonable size and doesn't have confidentiality problems?
? 
?   Ben Bolker
? 
? 	[[alternative HTML version deleted]]
? 

? _______________________________________________
? R-sig-mixed-models at r-project.org mailing list
? https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models

  
    

      
      
    

  


  


      

    

    

      
      

        
      

    

    

      
      

        
      

    

    

      
      

        


  

        

      Ben Bolker
    

    

      
      #
    

  


  

      
On 14-01-08 05:13 AM, Cole, Tim wrote:

            

      
      
      
    

        
In any case it still shouldn't crash with an obscure error message!
   I can imagine this issue *might* have been fixed by some of our
'clamping' fixes, i.e.

https://github.com/lme4/lme4/commit/c51a142be1781520f8bd1d7bb6aee27bf4e96568

  If you have a chance it would be great if you could test with a
development version of lme4 and see if the problem goes away ...

  Ben Bolker