glmm on paired data with repeated measures

classic Classic list List threaded Threaded
2 messages Options
Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

glmm on paired data with repeated measures

Mariano Devoto
Hi everyone. I'd really appreciate help with the following analysis.
Apologies for such a basic question; I looked for previous queries on
similar data, but none seem to quite fit what I am after.

I have a field experiment aimed at understanding how the presence of
"bodyguard" ants regulates the abundance of butterfly eggs and larvae (and
the damage the latter cause) on a focal plant species.
For this, I set up fourteen pairs of plants in a nature reserve and then
assigned each plant in each pair to one of two treatments, either "with
ants" or "without ants", by applying a physical barrier at the base of the
plant.
In the following nine weeks I measured four response variables once a week
on a subsample of ten randomly chosen leaves of each plant. I thus have
nine repeated measures on each plant.

My response variables are:
ants: number of ants (this was measured just to check the physical barrier
had worked OK)
eggs: number of butterfly eggs
larvae: number of butterfly larvae
dam: percent leaf damage (percentage eaten by larvae)

My explanatory variables are:
treat: treatment (two levels: "con" and "sin" mean with and without ants,
respectively)
pair: plant pair
date

For each response variable I would like to build a model that accounts for
the lack of independence of data within each pair, and that considers the
fact that data come from repeated measures (so, for instance, leaf damage
tends to accumulate).
My specific question is if including the random term "pair" in the model
accomplishes both things. I guess it probably doesn't, so I'd appreciate
any suggestions.
The results indicate plants without ants have a higher number of eggs
(which is what we expected, yeaeee), but without proper control of the
autocorrelations I mentioned I am not convinced.
I provide a workable example below.

#read data from Google drive
#each line represents the variables measured on a single leaf of a single
plant on a single date.

id <- "0Bzd8I1jr8z_iU0h6R0hxaElseDA" # google file ID
gaston <- read.table(sprintf("<a href="https://docs.google.com/uc?id=%s&export=">https://docs.google.com/uc?id=%s&export=
download", id), head=T)

#GLMM
require(lme4); require(effects)
M1 <- glmer(eggs ~ treat + (1|pair), data=gaston, family=poisson)

#check if there are any significant effects of treatment
summary(M1)
allEffects(M1)

Thanks in advance for your help.

Best,

Mariano

___
Dr. Mariano Devoto
Profesor Adjunto - Cátedra de Botánica General, Facultad de Agronomía de la
U.B.A.
Investigador Adjunto del CONICET

Av. San Martin 4453
C1417DSE
C.A. de Buenos Aires
Argentina

http://www.agro.uba.ar/users/mdevoto/

        [[alternative HTML version deleted]]

_______________________________________________
R-sig-ecology mailing list
[hidden email]
https://stat.ethz.ch/mailman/listinfo/r-sig-ecology
Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: glmm on paired data with repeated measures

Drew Tyre
I think you're pretty close. You want a factor that identifies the individual plant to get the repeated measures part correct, and I would add a fixed effect of days from the start to capture the trend over time (if there is one). So:


library(tidyverse)
library(lubridate)
gaston <- mutate(gaston,
                 plant = paste0(pair,treat),
                 date = dmy(date),
                 day = as.numeric((date - min(date))/86400))

ggplot(gaston, aes(x=day, y=eggs)) +
  geom_jitter(aes(shape=treat)) +
  facet_wrap(~pair)

M1 <- glmer(eggs ~ treat*day + (1|pair/plant), data=gaston, family=poisson)


And you should check for overdispersion. You might also check with r-sig-mixed-models.


--
Drew Tyre

School of Natural Resources
University of Nebraska-Lincoln
416 Hardin Hall, East Campus
3310 Holdrege Street
Lincoln, NE 68583-0974

phone: +1 402 472 4054
fax: +1 402 472 2946
email: [hidden email]
http://snr.unl.edu/tyre
http://atyre2.github.io

________________________________
From: R-sig-ecology <[hidden email]> on behalf of Mariano Devoto <[hidden email]>
Sent: Saturday, March 4, 2017 6:32:17 AM
To: [hidden email]
Subject: [R-sig-eco] glmm on paired data with repeated measures

Hi everyone. I'd really appreciate help with the following analysis.
Apologies for such a basic question; I looked for previous queries on
similar data, but none seem to quite fit what I am after.

I have a field experiment aimed at understanding how the presence of
"bodyguard" ants regulates the abundance of butterfly eggs and larvae (and
the damage the latter cause) on a focal plant species.
For this, I set up fourteen pairs of plants in a nature reserve and then
assigned each plant in each pair to one of two treatments, either "with
ants" or "without ants", by applying a physical barrier at the base of the
plant.
In the following nine weeks I measured four response variables once a week
on a subsample of ten randomly chosen leaves of each plant. I thus have
nine repeated measures on each plant.

My response variables are:
ants: number of ants (this was measured just to check the physical barrier
had worked OK)
eggs: number of butterfly eggs
larvae: number of butterfly larvae
dam: percent leaf damage (percentage eaten by larvae)

My explanatory variables are:
treat: treatment (two levels: "con" and "sin" mean with and without ants,
respectively)
pair: plant pair
date

For each response variable I would like to build a model that accounts for
the lack of independence of data within each pair, and that considers the
fact that data come from repeated measures (so, for instance, leaf damage
tends to accumulate).
My specific question is if including the random term "pair" in the model
accomplishes both things. I guess it probably doesn't, so I'd appreciate
any suggestions.
The results indicate plants without ants have a higher number of eggs
(which is what we expected, yeaeee), but without proper control of the
autocorrelations I mentioned I am not convinced.
I provide a workable example below.

#read data from Google drive
#each line represents the variables measured on a single leaf of a single
plant on a single date.

id <- "0Bzd8I1jr8z_iU0h6R0hxaElseDA" # google file ID
gaston <- read.table(sprintf("<a href="https://docs.google.com/uc?id=%s&export=">https://docs.google.com/uc?id=%s&export=
download", id), head=T)

#GLMM
require(lme4); require(effects)
M1 <- glmer(eggs ~ treat + (1|pair), data=gaston, family=poisson)

#check if there are any significant effects of treatment
summary(M1)
allEffects(M1)

Thanks in advance for your help.

Best,

Mariano

___
Dr. Mariano Devoto
Profesor Adjunto - C�tedra de Bot�nica General, Facultad de Agronom�a de la
U.B.A.
Investigador Adjunto del CONICET

Av. San Martin 4453
C1417DSE
C.A. de Buenos Aires
Argentina

http://www.agro.uba.ar/users/mdevoto/

        [[alternative HTML version deleted]]

_______________________________________________
R-sig-ecology mailing list
[hidden email]
https://stat.ethz.ch/mailman/listinfo/r-sig-ecology

        [[alternative HTML version deleted]]


_______________________________________________
R-sig-ecology mailing list
[hidden email]
https://stat.ethz.ch/mailman/listinfo/r-sig-ecology
Loading...