best choice of GLMM for seed set data

Dear Mariano,

The binomial distribution (not error family) assumes that you have a
number of successes and failures. If the potential number of seeds is
fixed by the morphology of the plant, then a binomial distribution is
reasonable. If the potential number of seeds is dictated by
morphology, then I'd rather see it as counts and use a Poisson or
negative binomial.

The correct syntax in the binomial case is cbind(success, failure). Or
in your case cbind(seeds, 4 - seeds).

Best regards,
ir. Thierry Onkelinx
Instituut voor natuur- en bosonderzoek / Research Institute for Nature
and Forest
team Biometrie & Kwaliteitszorg / team Biometrics & Quality Assurance
Kliniekstraat 25
1070 Anderlecht
Belgium

To call in the statistician after the experiment is done may be no
more than asking him to perform a post-mortem examination: he may be
able to say what the experiment died of. ~ Sir Ronald Aylmer Fisher
The plural of anecdote is not data. ~ Roger Brinner
The combination of some data and an aching desire for an answer does
not ensure that a reasonable answer can be extracted from a given body
of data. ~ John Tukey


2015-08-26 20:32 GMT+02:00 Mariano Devoto <mdevoto at agro.uba.ar>:

Dear all. I am analysing data from a field experiment on a crop
pollination. I want to test if there are differences in the number of seeds
per fruit between three treatments. The experimental design consists of
four separate sites where small subplots (ca. 5 plants each) received one
of the treatments. In each site, 8 subplots were allocated to treatment A,
8 to treatment B and 4 to treatment C. When fruits were ripe I collected
all plants from each subplot and counted stems, fruits per stem and seeds
per fruit. I think a GLMM is the best way to go as I expect random effects
related to field and subplot identity, and my response variable (number of
seeds) is clearly non-normal. My main concern is the choice of the error
family. As I?m counting seeds I first though of a Poisson model, but then
realized that seed numbers only range from 0 to 4. I am now considering
using a binomial model such as this:


glmer(cbind(seeds,4) ~ treatment + (1|site) + (1|subplot), data=seed.data,
family=binomial)


Does this make sense?


I would welcome any advice before hitting ?SEND? in Tinn-R :-).



--
*Mariano Devoto*

        [[alternative HTML version deleted]]

_______________________________________________
R-sig-ecology mailing list
R-sig-ecology at r-project.org
https://stat.ethz.ch/mailman/listinfo/r-sig-ecology