[R-sig-eco] Question about a resampling procedure with LME

Fri Nov 18 15:46:50 CET 2016

Hi Thiago,

I think you want a hierarchical model.

bird_richness ~ Poisson(mu_br)
log(mu_br) = Season + Pool

zooplankton ~ Poisson(mu_zp)
log(mu_zp) = Season + Pool + mu_br

I think that is is feasible to fit such model with the INLA package (
www.r-inla.org).

Best regards,

Thierry

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

2016-11-17 2:57 GMT+01:00 Thiago Gonçalves-Souza <tgoncalves.souza at gmail.com
>:

> Dear Thierry,
>
> Thank you for your remarks.
>
> I agree with you that I should use pool as a fixed effect.
>
> Let me explain some points you've raised:
>
> Bird diversity was obtained at the pool level. Unfortunately I don't have
> the coordinates of zooplankton (they're 50 cm apart). Only pool coordinates
> (n=2).
>
> Thank you again
>
> Best regards,
>
> 2016-11-16 7:01 GMT-02:00 Thierry Onkelinx <thierry.onkelinx at inbo.be>:
>
>> Dear Thiago,
>>
>> A few remarks.
>>
>> Two pools are not enough to get reliable estimates for the random effect.
>> Although it is conceptually a random effect, it is better to use it as a
>> fixed effect.
>>
>> At which level did you sample the bird diversity? I assume that it would
>> be much larger than 10 x 15 cm.  Is it a pool level or a smaller section of
>> the pool. Do you have coordinates of the sampling locations of both
>> zooplankton and the birds?
>>
>> Best regards,
>>
>> Thierry
>>
>> 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
>>
>> 2016-11-14 3:28 GMT+01:00 Thiago Gonçalves-Souza <
>> tgoncalves.souza at gmail.com>:
>>
>>> Hello all,
>>>
>>>
>>>
>>> First, I'm sorry for this long question, guys.
>>>
>>>
>>> I’d like to know if the diversity of waterbirds affect zooplankton
>>> diversity in natural temporary pools over two seasonal periods (wet and
>>> dry). I’ve sampled zooplankton in 38 10x15 cm plots on the pool’s shore
>>> sediment (in the same place where birds usually) in both seasonal
>>> periods.
>>> I used a model of bird feet and simulated them walking on the sediment of
>>> each plot. Then, I washed the bird feet (of each plot separately) into
>>> containers with filtered water. The zooplankton that emerged from this
>>> water were identified.
>>>
>>>
>>> Unfortunately, I was not able to collect birds at the same plot (and at
>>> the
>>> same time) that I simulated their footprint. Rather, I’ve made 38 samples
>>> (during 9 months immediately after each seasonal period) of bird
>>> communities (direct observation on each pool).
>>>
>>>
>>> Thus, I do not believe that can use any classic model (such as anova
>>> designs) because birds and zooplankton were not collected at the same
>>> sampling unit.
>>>
>>> To do this, I thought to randomly select *10* plots of each combination
>>> of
>>> zooplankton sampling, that is: 10 samples from pool 1/wet season, 10 from
>>> pool 1/dry season, 10 from pool 2/wet season and 10 from pool 2/dry
>>> season.
>>> I did the same with bird samples, selecting 10 samples by pool and season
>>> combination. Then, we run a lme model using species richness of birds
>>> (independent variable) and zooplankton (dependent variable), and season
>>> as
>>> independent variable. The variables were nested in pool. I repeated (for
>>> loop) this procedure 1000 times and calculated the number of times that
>>> the
>>> model was significant (p < 0.05) and the accompanying F values (for bird
>>> richness and season effects).
>>>
>>>
>>> I’d like to hear from you if this procedure make any sense. Please, see
>>> below a reproducible example. Thank you very much!
>>>
>>>
>>> Note: I’m pretty sure that the for loop could be shorter ;)
>>>
>>>
>>>
>>> library(nlme)
>>>
>>>
>>>
>>> #### Fake da ta
>>>
>>>
>>>
>>> ## zooplankton table
>>>
>>> ric.zoo <- rnorm(120, 20, 2)
>>>
>>> pool <- factor(rep(c("pool1", "pool2"), 60))
>>>
>>> season <- factor(rep(rep(c("dry", "wet"), each=30), 2))
>>>
>>> tab <- data.frame(pool, season, ric.zoo)
>>>
>>>
>>>
>>> ## birds table
>>>
>>> ric.bird <- rnorm(76, 10, 2)
>>>
>>> pool_b <- factor(rep(c("pool1", "pool2"), 38))
>>>
>>> season_b <- factor(rep(rep(c("dry", "wet"), each=19), 2))
>>>
>>> tab_b <- data.frame(pool=pool_b, season=season_b, ric.bird)
>>>
>>>
>>>
>>> ## Dataset to be used in for loops
>>>
>>> bird.p.val <- numeric()
>>>
>>> season.p.val <- numeric()
>>>
>>> bird.f <- numeric()
>>>
>>> season.f <- numeric()
>>>
>>> pool.sort <- rep(c("pool1", "pool2"), each=20)
>>>
>>> season.sort <- rep(rep(c("wet", "dry"), each=10), 2)
>>>
>>> factor.sort <- data.frame(pool=pool.sort, season=season.sort)
>>>
>>> nreps <- 1000
>>>
>>>
>>>
>>>
>>>
>>> for(i in 1:nreps){
>>>
>>>   subset1 <- subset(tab, pool=="pool1"&season=="wet")
>>>
>>>   subsample1 <- subset1[sample(nrow(subset1), 10),][,3]
>>>
>>>   subset2 <- subset(tab, pool=="pool1"&season=="dry")
>>>
>>>   subsample2 <- subset2[sample(nrow(subset2), 10),][,3]
>>>
>>>   subset3 <- subset(tab, pool=="pool2"&season=="wet")
>>>
>>>   subsample3 <- subset3[sample(nrow(subset3), 10),][,3]
>>>
>>>   subset4 <- subset(tab, pool=="pool2"&season=="dry")
>>>
>>>   subsample4 <- subset4[sample(nrow(subset4), 10),][,3]
>>>
>>>   subbird1 <- subset(tab_b, pool=="pool1"&season=="wet")
>>>
>>>   subsbird1 <- subbird1[sample(nrow(subbird1), 10),][,3]
>>>
>>>   subbird2 <- subset(tab_b, pool=="pool1"&season=="dry")
>>>
>>>   subsbird2 <- subbird2[sample(nrow(subbird2), 10),][,3]
>>>
>>>   subbird3 <- subset(tab_b, pool=="pool2"&season=="wet")
>>>
>>>   subsbird3 <- subbird3[sample(nrow(subbird3), 10),][,3]
>>>
>>>   subbird4 <- subset(tab_b, pool=="pool2"&season=="dry")
>>>
>>>   subsbird4 <- subbird4[sample(nrow(subbird4), 10),][,3]
>>>
>>>   ric.sort.zoo <- c(subsample1, subsample2, subsample3, subsample4)
>>>
>>>   ric.sort.birds <- c(subsbird1, subsbird2, subsbird3, subsbird4)
>>>
>>>   df40 <- data.frame(ric.zoo=ric.sort.zoo, ric.bird=ric.sort.birds,
>>> factor.sort)
>>>
>>>   lm.nested <- lme(ric.zoo~ric.bird+season, random=~1|pool, data=df40,
>>> method="REML")#MODELO
>>>
>>>   mod <- anova.lme(lm.nested, type="sequential", adjustSigma = FALSE)
>>>
>>>   bird.p.val[i] <- mod$`p-value`[2]
>>>
>>>   season.p.val[i] <- mod$`p-value`[3]
>>>
>>>   bird.f[i] <- mod$`F-value`[2]
>>>
>>>   season.f[i] <- mod$`F-value`[3]
>>>
>>> }
>>>
>>>
>>>
>>>
>>>
>>> sum(bird.p.val<0.05 )/nreps
>>>
>>> sum(season.p.val<0.05)/nreps
>>>
>>> --
>>>
>>> *Thiago Gonçalves-SouzaProfessor Adjunto*
>>>
>>> *Universidade Federal Rural de Pernambuco (UFRPE)*
>>> *Departamento de Biologia / Área de Ecologia *
>>>
>>> *E-mail alternativo: thiagoaracno at gmail.com <thiagoaracno at gmail.com>Home
>>> page:<http://thiagocalvesouza.wix.com/ecoffun
>>> <http://thiagocalvesouza.wix.com/ecoffun>**>*
>>> *Google Scholar:
>>> <http://scholar.google.com.br/citations?user=TjaP2l8AAAAJ&hl=pt-BR
>>> <http://scholar.google.com.br/citations?user=TjaP2l8AAAAJ&hl=pt-BR>>*
>>>
>>>         [[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
>>
>>
>>
>
>
> --
>
> *Thiago Gonçalves-SouzaProfessor Adjunto*
>
> *Universidade Federal Rural de Pernambuco (UFRPE)*
> *Departamento de Biologia / Área de Ecologia *
>
> *E-mail alternativo: thiagoaracno at gmail.com <thiagoaracno at gmail.com>Home
> page:<http://thiagocalvesouza.wix.com/ecoffun
> <http://thiagocalvesouza.wix.com/ecoffun>**>*
> *Google Scholar:
> <http://scholar.google.com.br/citations?user=TjaP2l8AAAAJ&hl=pt-BR
> <http://scholar.google.com.br/citations?user=TjaP2l8AAAAJ&hl=pt-BR>>*
>

	[[alternative HTML version deleted]]