Dear R-Users,
I have a question to the GLMM via the lme4 package.
I have 160 nest-boxes which are placed at 8 different localities.
"Count" is the number of animals which were found inside the boxes
during the observation time.
The independent variables are factors which are supposed to influence
the occurence of the animal.
1. Is the formula right like this?
Here are my outcomes of running the glmer function:
m1<-glmer(Count~A1+A2+A3+A4+A5+(1|locality), family=poisson(log))
summary(m1)
Generalized linear mixed model fit by the Laplace approximation
Formula: Count ~ A1 + A2 + A3 + A5 + A5 + (1 | locality)
AIC BIC logLik deviance
477.8 499.4 -231.9 463.8
Random effects:
Groups Name Variance Std.Dev.
locality (Intercept) 0.15181 0.38963
Number of obs: 160, groups: locality, 8
Fixed effects:
Estimate Std. Error z value Pr(>|z|)
(Intercept) -4.410558 1.854530 -2.378 0.017394 *
A1 -0.003627 0.002199 -1.649 0.099089 .
A2 0.037399 0.018599 2.011 0.044350 *
A3 0.041520 0.018579 2.235 0.025432 *
A4 0.115765 0.034663 3.340 0.000839 ***
A5 0.532708 0.073443 7.253 4.06e-13 ***
---
Signif. codes: 0 ?***? 0.001 ?**? 0.01 ?*? 0.05 ?.? 0.1 ? ? 1
Correlation of Fixed Effects:
(Intr) S2 K2 SP2 SUMSP2
A1 0.005
A2 -0.989 -0.030
A3 -0.984 -0.123 0.992
A4 0.001 -0.076 -0.068 -0.036
A5 -0.037 0.123 -0.058 -0.060 -0.041
2. Does it make sense to manually cut out these values which are not significant
and to it compare the AIC's of the different models?
There are still more variables which I didn't took inside to keep the
example little bit shorter.
Because of an advise I wanted to build up the model by stepwise regression.
I actually wanted to use the functions drop1/add1 which are not possible for
this object.
Thanks a lot for every help!
Kind regards,
Jan Kaesler
--
Aufgepasst: Sind Ihre Daten beim Online-Banking auch optimal gesch?tzt?
Jetzt absichern: https://homebanking.gmx.net/?mc=mail at footer.hb