Dear users,
Thanks for your attention. I’m running a glmm model using the glmmadmb function
provided in the package glmmADMB.
My dependent variable is the number of individuals belonging to a single
species of an aquatic insect, sampled throughout two non-consecutive years. The
samples were classified by the following fixed factors:
- year: two levels (2004, 2009);
- hydroperiod (hyd): classified in two levels (high and low flow);
- daytime (time): two levels, night or day;
- stratification (str): two levels, bottom and surface
- water current velocity (vel): quantitative variable used as an offset, since
the sampling method is very sensitive for the amount of water filtered, which
has a strong correlation with water current velocity.
A single random term was added to the model, named as sampleID, since sampling
at the bottom and at the surface were performed at the same moment (as far as
understand, the inclusion of such random factor will treta them as a sampling
block). I also added two interaction terms (hydroperiod:daytime,
hydroperiod:stratification).
The model that I tested was a confirmatory one, based on a very precise
biological hypothesis, resulting in the following output:
-----------
glmmadmb(formula = Count ~ year + hyd * (time + str) + offset(vel) + (1 |
sampleID),
family = “nbinom”, zeroInflation = T)
AIC: 1484.1
Coefficients:
Estimate Std. Error z value Pr(>|z|)
(Intercept) -0.339 0.229 -1.48 0.13756
year2009 -0.164 0.148 -1.11 0.26852
hydlow 0.197 0.259 0.76 0.44747
timenight -0.556 0.254 -2.19 0.02851 *
strsurf 0.808 0.215 3.75 0.00017 ***
hydlow:timenight 0.709 0.311 2.28 0.02270 *
hydlow:strsurf -0.195 0.263 -0.74 0.45832
—
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
Number of observations: total=366, sample=183
Random effect variance(s):
Group=sampleID
Variance StdDev
(Intercept) 0.2813 0.5304
Negative binomial dispersion parameter: 1.3265 (std. err.: 0.25595)
Zero-inflation: 1.0003e-06 (std. err.: 6.2823e-06 )
Log-likelihood: -732.046
---------------
One graphical output that I opted to use shows the estimates provided by the
model and their respective confidence intervals (I used the coefplot2 function).
Now I’m (desperately) trying to provide a better graphical representation about
the predicted values from the model, in order to express graphically the
magnitude and direction of variation explained by the model. However, I’m not
sure if the data that I should use for such description comes from the
following indexation
model$fitted
or if I could use the command:
plot(interactionMeans(model))
Thank you so much for your attention,
Tchê
--
Luiz Ernesto Costa-Schmidt
http://lattes.cnpq.br/1402956553786728 <http://lattes.cnpq.br/1402956553786728>
Pós-doutorando - PNPD/CAPES
Universidade do Vale do Rio dos Sinos - UNISINOS
Programa de Pós-Graduação em Biologia
Avenida Unisinos, 950 - Sala E04 235
CEP 93022-000
São Leopoldo/RS - Brasil
Telefone: +55 51 3590.8477
http://www.unisinos.br/mestrado-e-doutorado/biologia
<http://www.unisinos.br/mestrado-e-doutorado/biologia>
[[alternative HTML version deleted]]
______________________________________________
R-help@r-project.org mailing list -- To UNSUBSCRIBE and more, see
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
