multivariate mixed nested model

Hi Claudio, for multivariate data see the mvabund package as well as the boral package.
Regards,
Ludovico

Ottieni Outlook per Android<https://aka.ms/ghei36>


Da: Claudio
Inviato: sabato 28 gennaio, 16:45
Oggetto: [R-sig-ME] multivariate mixed nested model
A: r-sig-mixed-models at r-project.org

Hi all. I collected six body features (bf1-bf6)from three populations of a salamander and from two populations of another sister species of salamander. I would evaluate how the species (fixed) and population belonging (random) affect the body features, by comparing models built with lme4. For some models, I also want to include bf6 as covariate. Thus, in case of univariate analyses, some models, for example, could be: mo1

Hi Ludovico,
thanks for your suggestion. However, it seems that they both deal with
abundance data, while I measured body features.
best regards
Claudio

Il giorno lun, 30/01/2017 alle 06.22 +0000, Ludovico Frate ha scritto:

Hi Claudio, for multivariate data see the mvabund package as well as
the boral package.
Regards,
Ludovico

Ottieni Outlook per Android


Da: Claudio
Inviato: sabato 28 gennaio, 16:45
Oggetto: [R-sig-ME] multivariate mixed nested model
A: r-sig-mixed-models at r-project.org
Hi all. I collected six body features (bf1-bf6)from three populations
of a salamander and from two populations of another sister species of
salamander. I would evaluate how the species (fixed) and population
belonging (random) affect the body features, by comparing models
built with lme4. For some models, I also want to include bf6 as
covariate. Thus, in case of univariate analyses, some models, for
example, could be: mo1

Hi Claudio,

both packages can deal with different family, i.e. poisson and negative binomial for counts, beta for proportion, normal...

Ragards,

Ludovico



Dott. For. Ludovico Frate, Ph.D.

Environmetrics Lab,
Dipartimento di Bioscienze e Territorio - Universit? degli Studi del Molise.
Contrada Fonte Lappone, 86090 - Pesche (IS) - ITALIA.
Cel: ++39 3333767557|Fax: ++39 (0874) 404123
E-mail: frateludovico at gmail.com<mailto:frateludovico at gmail.com>|ludovicofrate at hotmail.it<mailto:ludovicofrate at hotmail.it>

Inviato: luned? 30 gennaio 2017 15.49
A: Ludovico Frate; r-sig-mixed-models at r-project.org
Oggetto: Re: [R-sig-ME] multivariate mixed nested model

Hi Ludovico,
thanks for your suggestion. However, it seems that they both deal with
abundance data, while I measured body features.
best regards
Claudio

Il giorno lun, 30/01/2017 alle 06.22 +0000, Ludovico Frate ha scritto:
> Hi Claudio, for multivariate data see the mvabund package as well as
> the boral package.
> Regards,
> Ludovico
>
> Ottieni Outlook per Android
>
>
> Da: Claudio
> Inviato: sabato 28 gennaio, 16:45
> Oggetto: [R-sig-ME] multivariate mixed nested model
> A: r-sig-mixed-models at r-project.org
> Hi all. I collected six body features (bf1-bf6)from three populations
> of a salamander and from two populations of another sister species of
> salamander. I would evaluate how the species (fixed) and population
> belonging (random) affect the body features, by comparing models
> built with lme4. For some models, I also want to include bf6 as
> covariate. Thus, in case of univariate analyses, some models, for
> example, could be: mo1
>

Dear Claudio,

I this you need to add the interaction with traits to all the fixed and
random effects. Otherwise you assume that these have the same effect for
each trait. Note that 0 + traits is identical to traits - 1.

mumo1 <- lmer(value~0 + traits + traits:species + (0 +
traits|species:populations) + (0 + traits|individuals), mdata, REML=FALSE)

Your second question needs a reproducible example.

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

2017-01-28 16:43 GMT+01:00 Claudio <oppela at gmail.com>:

Hi all.
I collected six body features (bf1-bf6)from three populations of a
salamander and from two populations of another sister species of
salamander.
I would evaluate how the species (fixed) and population belonging
(random) affect the body features, by comparing models built with lme4.
For some models, I also want to include bf6 as covariate. Thus, in case
of univariate analyses, some models, for example, could be:
mo1<-lmer(bf1~species+(1|species:population), data, REML=FALSE)
mo2<-lmer(bf1~species+bf6+(1|species:population), data, REML=FALSE)

However, I want to fit multivariate models, and my post is about this.
First, I melted the data:
mdata<-melt(data, id.vars = c("species", "population", "bf6"),
measure.vars = c("bf1", "bf2","bf3","bf4","bf5"), variable.name =
"traits)

Now the question.
1) Are the multivariate versions of the models mo1 and mo2 above
mumo1<-lmer(value~traits -1 + species + (1|species:populations) +
(1|individuals), mdata, REML=FALSE)
mumo1<-lmer(value~traits -1 + species + bf6 + (1|species:populations) +
(1|individuals), mdata, REML=FALSE)

A secondary question, which in case I will move to a new post:
it seemed to me that building multivariate models with MCMCglmm is
easier. However, cbind did not work, even without missing values: to
your knowledge, is there any issue?

thanks in advance
Claudio

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