R help - Oct 2008 - false convergence (8) after removal of the two-way interaction

Dear All,

I am working with a generalized linear mixed-effects model with poisson error 
using the lme4 package in R. I created a model with the lmer function including
some main effects, three two way interactions and two random effects.
The model works well, but I have troubles when removing on of the two-way
interactions. The Warning message: "In mer_finalize(ans) : false
convergence (8)" appears.

I looked for advice on this error message in the R-help, but could not find the
answer as to why this happens.

Both variables in the interaction are are continuous with a normal distribution.
The model looks like following model: 
year=factor(year)
nk=factor(nk)
rnr=factor(rnr)
VAR1-5 are normally distributed continuous variables

model1<-lmer(recruit~(year:VAR1)+year+VAR1+VAR2+VAR3+VAR4+
VAR5+(1|nk)+(1|rnr),family=poisson)

Generalized linear mixed model fit by the Laplace approximation 
recruit ~ (year:VAR1) + year + VAR1+ VAR2 + VAR3 + VAR4 + VAR5 + (1 | nk) + (1 |
rnr)
  AIC   BIC     logLik deviance
 306.9 394.8  -132.5    264.9
Random effects:
 Groups Name        Variance Std.Dev.
 rnr    (Intercept) 0.00000  0.00000 
 nk     (Intercept) 0.46014  0.67834 
Number of obs: 486, groups: rnr, 341; nk, 125

Fixed effects:
                          Estimate Std. Error z value Pr(>|z|)   
(Intercept)         -1.513e-01  1.356e+00 -0.1116  0.91112   
VAR2   		-1.440e-02  6.429e-02 -0.2241  0.82271   
VAR3	              3.130e-04  1.413e-03  0.2215  0.82469   
year1              -2.671e-01  5.719e-01 -0.4670  0.64050   
year2               4.888e-01  4.792e-01  1.0199  0.30780   
year3              -2.720e+00  1.169e+00 -2.3260  0.02002 * 
year4               1.287e+00  4.211e-01  3.0569  0.00224 **
year5              -1.180e-01  4.195e-01 -0.2812  0.77855   
year6               7.367e-01  4.249e-01  1.7340  0.08292 . 
VAR1    	-2.083e-02  4.826e-02 -0.4317  0.66596   
VAR4                  4.022e-03  3.500e-03  1.1491  0.25050   
VAR5                 -5.274e-02  2.622e-02 -2.0114  0.04428 * 
year1:VAR1    -1.504e-01  1.193e-01 -1.2609  0.20735   
year2:VAR1      5.708e-02  6.833e-02  0.8353  0.40354   
year3:VAR1     -6.112e-01  2.824e-01 -2.1645  0.03043 * 
year4:VAR1      2.737e-03  6.777e-02  0.0404  0.96778   
year5:VAR1      4.702e-02  8.663e-02  0.5428  0.58729   
year6:VAR1      8.853e-02  7.870e-02  1.1249  0.26062   


Model2<-lmer(recruit~year+VAR1+VAR2+VAR3+VAR4+VAR5+(1|nk)+(1|rnr),family=poisson)
After removal of the interaction the false convergence appears.
Could anybody give me advice on how to solve this problem, please? 

Thank you in advance,
Eva Fucikova

	[[alternative HTML version deleted]]