[R-sig-ME] False convergence in null model

[Iker Vaquero Alba]

   Dear list:

   Sometimes I have had "false convergence" problems while performing some analyses with lmer. Usually, I was told that the problem could be an overparameterization. But this time, I've got the false convergence message just at the end of a stepwise model simplification, after removing the last remaining term. Obviously, I would like to know the reason for this, and I would appreciate any ideas about it. I attach the data set I am using. The analysis was performed after standardizing the continuous independent variables (tlength, brithr, briventral, inslarge and cond). 

   This is the step where I get the error message:

   s.clutchmodel1.57<-update(s.clutchmodel1.56,~.-brithrst)
Mensajes de aviso perdidos
In mer_finalize(ans) : singular convergence (7)
> summary(s.clutchmodel1.57)
Generalized linear mixed model fit by the Laplace approximation 
Formula: clsize1 ~ (1 | site/pair) + (1 | year) 
 AIC   BIC logLik deviance
  30 39.68    -11       22
Random effects:
 Groups    Name        Variance Std.Dev.
 pair:site (Intercept)  0        0      
 site      (Intercept)  0        0      
 year      (Intercept)  0        0      
Number of obs: 83, groups: pair:site, 37; site, 12; year, 2

Fixed effects:
            Estimate Std. Error z value Pr(>|z|)    
(Intercept)  1.60702    0.04915    32.7   <2e-16 ***
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 
> summary(s.clutchmodel1.56)
Generalized linear mixed model fit by the Laplace approximation 
Formula: clsize1 ~ brithrst + (1 | site/pair) + (1 | year) 
   AIC   BIC logLik deviance
 31.52 43.62 -10.76    21.52
Random effects:
 Groups    Name        Variance   Std.Dev.  
 pair:site (Intercept) 0.0000e+00 0.0000e+00
 site      (Intercept) 7.6128e-17 8.7251e-09
 year      (Intercept) 1.0638e-14 1.0314e-07
Number of obs: 83, groups: pair:site, 37; site, 12; year, 2

Fixed effects:
            Estimate Std. Error z value Pr(>|z|)    
(Intercept)  1.60645    0.04918   32.67   <2e-16 ***
brithrst     0.03418    0.04930    0.69    0.488    
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 

Correlation of Fixed Effects:
         (Intr)
brithrst -0.034



   Do you think the problem could be in the fact that I am including random factors which can explain none of the variance? Should I do the analysis again without including those factors?

   Thank you very much.

   Iker 
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: Clutch Size1 2009+2010.txt
URL: <https://stat.ethz.ch/pipermail/r-sig-mixed-models/attachments/20110823/fffe3b54/attachment-0003.txt>