R help - May 2013 - Post hoc test for GLM with poisson distribution

Hi R-people,

 I performed controlled experiments to evaluated the seeds germination of
two palms under four levels of water treatments. I conducted a generalized
linear model (GLM) with a Poisson distribution to verify whether there were
significant differences in the number of seed germination (NS-count
variable)  between treatments and species (explanatory variables). Thus, my
model and output were:

model1<-glm(NS~Treatments*Species, family="poisson")



Coefficients:

                               Estimate  Std. Error  z value  Pr(>|z|)

(Intercept)               2.56247    0.57544   4.453    8.46e-06 ***

Treatments              -2.07267    0.35065  -5.911   3.40e-09 ***

Species                     -0.00312    0.30527  -0.010  0.992

Treatments:Species     0.90397    0.17896   5.051  4.39e-07 ***

 Null deviance: 379.870  on 98  degrees of freedom

Residual deviance:  68.302  on 95  degrees of freedom



There is a significant interaction between treatments:species. Which is the
post hoc test appropriate for this model?



Thanks,

Maria Isabel

	[[alternative HTML version deleted]]

On May 14, 2013, at 21:04 , Bel Braz wrote:
> Hi R-people,
> 
> I performed controlled experiments to evaluated the seeds germination of
> two palms under four levels of water treatments. I conducted a generalized
> linear model (GLM) with a Poisson distribution to verify whether there were
> significant differences in the number of seed germination (NS-count
> variable)  between treatments and species (explanatory variables). Thus, my
> model and output were:
> 
> model1<-glm(NS~Treatments*Species, family="poisson")
> 
> 
> 
> Coefficients:
> 
>                               Estimate  Std. Error  z value  Pr(>|z|)
> 
> (Intercept)               2.56247    0.57544   4.453    8.46e-06 ***
> 
> Treatments              -2.07267    0.35065  -5.911   3.40e-09 ***
> 
> Species                     -0.00312    0.30527  -0.010  0.992
> 
> Treatments:Species     0.90397    0.17896   5.051  4.39e-07 ***
> 
> Null deviance: 379.870  on 98  degrees of freedom
> 
> Residual deviance:  68.302  on 95  degrees of freedom
> 
> 
> 
> There is a significant interaction between treatments:species. Which is the
> post hoc test appropriate for this model?
There's not much post hoc testing to do if the effect is described by a
single coefficient. Did you forget to code Treatments as a factor variable?

Apart from that, it depends on what you want to do. Do you want to know where
the interaction comes from, or just within which treatment(s) there is a species
effect?

Since there is only two species, the easiest way forward is to compute the four
species effects, one for each treatment. You can then compare the effects
pairwise (6 comparisons)  or compare each effect to zero (4 comparison). I
don't think you can do much better than simple Bonferroni corrections in
either case.
> 
> 
> 
> Thanks,
> 
> Maria Isabel
> 
> 	[[alternative HTML version deleted]]
> 
> ______________________________________________
> R-help at r-project.org mailing list
> 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.
-- 
Peter Dalgaard, Professor,
Center for Statistics, Copenhagen Business School
Solbjerg Plads 3, 2000 Frederiksberg, Denmark
Phone: (+45)38153501
Email: pd.mes at cbs.dk  Priv: PDalgd at gmail.com