[R] LMER quasibinomial

[Daniel Malter]

Hi,

a while ago I posted a question regarding the use of alternative models,
including a quasibinomial mixed-effects model (see Results 1). I rerun the
exact same model yesterday using R 2.7.2 and lme4_0.999375-26 (see Results
2) and today using R 2.7.2 and lme4_0.999375-27 (see Results 3). 

While the coefficient estimates are basically the same in all three
regressions, the estimated standard errors and t-values vary dramatically
(also the variance of the random effects) despite running the exact same
model. Is there any advice which of the models to trust and as to where
these differences come from?

Thanks,
Daniel

---------
Results 1
---------
Generalized linear mixed model fit using Laplace 
Formula: prob.bind ~ capacity * group + (1 | subject) 
 Subset: c(combination == "gnl") 
 Family: quasibinomial(logit link)
   AIC   BIC logLik deviance
 11082 11109  -5534    11068
Random effects:
 Groups   Name        Variance Std.Dev.
 subject  (Intercept) 42.977   6.5557  
 Residual             26.845   5.1813  
number of obs: 360, groups: subject, 90

Fixed effects:
                Estimate Std. Error t value
(Intercept)      -3.8628     1.2701  -3.041
capacity          1.1219     0.1176   9.542
group2            0.9086     1.7905   0.507
group3            2.3700     1.7936   1.321
capacity:group2  -0.1745     0.1610  -1.083
capacity:group3  -0.3807     0.1622  -2.348

Correlation of Fixed Effects:
            (Intr) capcty group2 group3 cpct:2
capacity    -0.322                            
group2      -0.709  0.228                     
group3      -0.708  0.228  0.502              
capcty:grp2  0.235 -0.730 -0.310 -0.167       
capcty:grp3  0.233 -0.725 -0.166 -0.305  0.529 


---------
Results 2
---------
Generalized linear mixed model fit by the Laplace approximation 
Formula: entryprob.bind ~ capacity * factor(group) + (1 | subject) 
   Data: res 
 Subset: c(which(is.na(entryprob) == FALSE) & combination == "gnl") 
   AIC   BIC logLik deviance
 11084 11115  -5534    11068
Random effects:
 Groups   Name        Variance Std.Dev.
 subject  (Intercept) 0.021575 0.14688 
 Residual             0.013457 0.11601 
Number of obs: 360, groups: subject, 90

Fixed effects:
                         Estimate Std. Error t value
(Intercept)             -3.864981   0.028454  -135.8
capacity                 1.121713   0.002632   426.1
factor(group)2           0.909167   0.040113    22.7
factor(group)3           2.372638   0.040184    59.0
capacity:factor(group)2 -0.173956   0.003606   -48.2
capacity:factor(group)3 -0.380799   0.003631  -104.9

Correlation of Fixed Effects:
            (Intr) capcty fct()2 fct()3 cp:()2
capacity    -0.322                            
factr(grp)2 -0.709  0.228                     
factr(grp)3 -0.708  0.228  0.502              
cpcty:fc()2  0.235 -0.730 -0.309 -0.166       
cpcty:fc()3  0.233 -0.725 -0.165 -0.304  0.529


-----------
Results 3
-----------
Generalized linear mixed model fit by the Laplace approximation 
Formula: entryprob.bind ~ capacity * factor(group) + (1 | subject) 
   Data: res 
 Subset: c(which(is.na(entryprob) == FALSE) & combination == "gnl") 
   AIC   BIC logLik deviance
 11082 11109  -5534    11068
Random effects:
 Groups  Name        Variance Std.Dev.
 subject (Intercept) 1.6032   1.2662  
Number of obs: 360, groups: subject, 90

Fixed effects:
                        Estimate Std. Error z value Pr(>|z|)    
(Intercept)             -3.86498    0.24528  -15.76  < 2e-16 ***
capacity                 1.12171    0.02269   49.43  < 2e-16 ***
factor(group)2           0.90917    0.34578    2.63  0.00856 ** 
factor(group)3           2.37264    0.34639    6.85 7.41e-12 ***
capacity:factor(group)2 -0.17396    0.03108   -5.60 2.18e-08 ***
capacity:factor(group)3 -0.38080    0.03130  -12.17  < 2e-16 ***
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 

Correlation of Fixed Effects:
            (Intr) capcty fct()2 fct()3 cp:()2
capacity    -0.322                            
factr(grp)2 -0.709  0.228                     
factr(grp)3 -0.708  0.228  0.502              
cpcty:fc()2  0.235 -0.730 -0.309 -0.166       
cpcty:fc()3  0.233 -0.725 -0.165 -0.304  0.529


-------------------------
cuncta stricte discussurus