[R] random effects in lmer
Line Johansen
line.johansen at bio.ntnu.no
Fri Aug 22 14:09:43 CEST 2008
Dear all<>
I am running several generalized mixed model using lmer. <>The models
typical look like this:
model2xx<-lmer(numbers~Treatment+b+(1|Genotype)+(1|Field)+(1|Genotype:Treatment),
family=quasipoisson)<>
All factors are categorical.
<>And the output looks like this:
Generalized linear mixed model fit using Laplace
formula:numbers~Treatment+Dead+(1|Genotype)+(1|Field)+(1|Genotype:Treatment),
family=quasipoisson)
Family: quasipoisson(log link)
AIC BIC logLik deviance
1033 1051 -509.4 1019
<>Random effects:
Groups Name Variance Std.Dev.
Genotype:Treatment (Intercept) 0.40555 0.63683
Genotype (Intercept) 1.16642 1.08001
Field (Intercept) 1.23738 1.11238
Residual 9.47740 3.07854
number of obs: 100, groups: Genotype1:Treatment1, 14; Genotype1, 5;
Felt1, 4
Fixed effects:
<>Estimate Std. Error t value
(Intercept) 3.20061 0.80010 4.000
Treatment a -0.05482 0.42619 -0.129
Treatment c 0.08316 0.46395 0.179
Dead No 0.27604 0.14873 1.856
Correlation of Fixed Effects:<>
(Intr) Tretmnt1 Trtmnt1c
Treatment a -0.247
Treatment c -0.239 0.450
Dead No -0.111 -0.132 0.003
<>
I need some help to evaluate the importance of the random factors. The
random factor of interest is Genotype. I have tried to delete random
factors from the model and comparing the model with the original model
by log likelihood-ratio statistics. Is this an appropriate method for
testing the random factors in lmer? Is it possible to evaluate how much
of the total variation the random factor Genotype explains?
<>
I have am a new user in lmer and my questions is probably very naive.
But I appreciate any help.
Thanks for the help. <>
Line
--
****************************************************************************
Line Johansen
Department of Biology
Norwegian University of Natural Science and Technology
Høgskoleringen 15
No-7491 Trondheim
Phone: 73 55 12 94
More information about the R-help
mailing list