[R] Help on estimated variance in lme4

[jerome lemaitre]

Dear all,

I get an error message when I run my model and I am not sure what to do
about it.

I try to determine what factors influence the survival of voles. I use a
mixed-model because I have several voles per site (varying from 2 to 19
voles).

Here is the model:
###
fm5 <-lmer(data=cdrgsaou2,
alive~factor(pacut)+factor(agecamp)+factor(sex)+ResCondCorp+(1|factor(cdrgsa
ou2$ids)),
	family=binomial,
	method="Laplace",
	)
###
Description of variables
Alive: 0 or 1; dead or alive
pacut: 0 or 1; presence of parasites
agecamp: a or j; adult or juvenile
sex: m or f; male or female
ResCondCorp: body condition, continuous;
cdrgsaou2$ids: name of the site.


Here is the output:

###
Generalized linear mixed model fit using Laplace 
Formula: alive ~ factor(pacut) + factor(agecamp) + factor(sex) + ResCondCorp
+      (1 | factor(cdrgsaou2$ids)) 
   Data: cdrgsaou2 
 Family: binomial(logit link)
      AIC      BIC    logLik deviance
 305.7418 328.7331 -146.8709 293.7418
Random effects:
 Groups                Name        Variance Std.Dev.
 factor(cdrgsaou2$ids) (Intercept) 0.034382 0.18542 
number of obs: 341, groups: factor(cdrgsaou2$ids), 36

Estimated scale (compare to 1)  2.174681 

Fixed effects:
                  Estimate Std. Error z value  Pr(>|z|)    
(Intercept)       0.971458   0.250951  3.8711 0.0001083 ***
factor(pacut)1   -0.831888   0.358583 -2.3199 0.0203447 *  
factor(agecamp)j -1.294236   0.330638 -3.9144 9.065e-05 ***
factor(sex)m      0.581713   0.296229  1.9637 0.0495616 *  
ResCondCorp      -0.176251   0.020263 -8.6982 < 2.2e-16 ***
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 

Correlation of Fixed Effects:
            (Intr) fct()1 fctr(g) fctr(s)
factr(pct)1 -0.334                       
fctr(gcmp)j -0.417  0.066                
factor(sx)m -0.505 -0.002 -0.173         
ResCondCorp -0.309 -0.010  0.302  -0.032 
###

Here is the error message:

###
Warning message:
Estimated variance for factor 'factor(cdrgsaou2$ids)' is effectively zero
 in: LMEopt(x = mer, value = cv)
###

Thank you very much by advance for any help.



Jérôme Lemaître


Ph.D. student
Silviculture-wildlife research chair in irregular boreal forests
& Départment of biology,
Faculty of Sciences and Engineering
Alexandre-Vachon building
University Laval
Quebec, QC  G1K 7P4
Phone : (418) 656-2131 poste 2917
Office : VCH-2044
Email: jerome.lemaitre.1 at ulaval.ca