simMVRM: The simulated multivariate repeated measures "simMVRM"...
In DImodelsMulti: Fit Multivariate Diversity-Interactions Models with Repeated Measures
simMVRM R Documentation
The simulated multivariate repeated measures "simMVRM" dataset
Description
The simMVRM dataset was simulated from a multivariate repeated measures DI model. It contains
336 plots comprising of four species that vary in proportions (p1 - p4). There are
three simulated responses (Y1, Y2, Y3), taken at two differing time points, recorded in a
wide data format (one column per response type). The data was simulated assuming that there were
existing covariances between the responses and between the time pointsand both species identity and
species interaction effects were present.
Usage
data("simMVRM")
Format
A data frame with 672 observations on the following 9 variables.
plota factor vector indicating the ID of the experimental unit
p1a numeric vector indicating the initial proportion of species 1
p2a numeric vector indicating the initial proportion of species 2
p3a numeric vector indicating the initial proportion of species 3
p4a numeric vector indicating the initial proportion of species 4
Y1a numeric vector indicating the response of ecosystem function 1
Y2a numeric vector indicating the response of ecosystem function 2
Y3a numeric vector indicating the response of ecosystem function 3
timea factor with levels 1 2
Details
What are Diversity-Interactions (DI) models?
Diversity-Interactions (DI) models (Kirwan et al., 2009) are a set of tools for analysing and
interpreting data from experiments that explore the effects of species diversity on community-level
responses. We strongly recommend that users read the short introduction to Diversity-Interactions
models (available at: DImodels). Further information on
Diversity-Interactions models is also available in Kirwan et al., 2009 and
Connolly et al., 2013.
Parameter values for the simulation
Multivariate repeated measures DI models take the general form of:
{y}_{kmt} = {Identities}_{kmt} + {Interactions}_{kmt} + {Structures}_{kt} + {\epsilon}_{kmt}
where y are the community-level responses, the Identities are the effects of species
identities for each response and enter the model as individual species proportions measured at the
beginning of the time period, the Interactions are the interactions among the species
proportions, while Structures include other experimental structures such as blocks,
treatments, or density.
The dataset simRM was simulated with:
identity effects for the four species for each time and ecosystem function:
Y1time1 = -1.0, 5.0, 2.8, -0.9
Y1time2 = 0.5, 5.4, 4.9, -2.1
Y2time1 = 0.1, 4.1, -0.5, 0.3
Y2time2 = 2.3, 3.2, -3.1, 2.1
Y3time1 = 0.9, 6.6, 3.5, 6.1
Y3time2 = -0.1, 7.0, 2.8, 4.0
evenness interaction effect for each time and ecosystem function:
Y1time1 = -0.1
Y1time2 = 12.0
Y2time1 = 2.3
Y2time2 = 1.6
Y3time1 = 2.1
Y3time2 = 6.8
-
\epsilon assumed to have a multivaraite normal distribution with mean 0. An ecosystem
function matrix Sigma:
[1,] 2.36 0.71 -0.29
[2,] 0.71 1.42 -0.39
[3,] -0.29 -0.39 2.21
and a time matrix Sigma:
[1,] 1.69 0.46
[2,] 0.46 1.31
References
Dooley, A., Isbell, F., Kirwan, L., Connolly, J., Finn, J.A. and Brophy, C., 2015.
Testing the effects of diversity on ecosystem multifunctionality using a multivariate model.
Ecology Letters, 18(11), pp.1242-1251.
Finn, J.A., Kirwan, L., Connolly, J., Sebastia, M.T., Helgadottir, A., Baadshaug, O.H.,
Belanger, G., Black, A., Brophy, C., Collins, R.P. and Cop, J., 2013.
Ecosystem function enhanced by combining four functional types of plant species in intensively
managed grassland mixtures: a 3-year continental-scale field experiment.
Journal of Applied Ecology, 50(2), pp.365-375 .
Kirwan, L., Connolly, J., Finn, J.A., Brophy, C., Luscher, A., Nyfeler, D. and Sebastia, M.T.,
2009.
Diversity-interaction modeling: estimating contributions of species identities and interactions
to ecosystem function.
Ecology, 90(8), pp.2032-2038.
Examples
###################################################################################################
###################################################################################################
## Modelling Example
# For a more thorough example of the workflow of this package, please see vignette
# DImulti_workflow using the following code:
# vignette("DImulti_workflow")
head(simMVRM)
# We call DImulti() to fit a series of models, with increasing complexity, and test whether the
# additional terms are worth keeping.
# We begin with an ID DI model, ensuring to use method = "ML" as we will be comparing fixed effects
MVRMmodel <- DImulti(y = 6:8, eco_func = c("Na", "un"), time = c("time", "CS"), unit_IDs = 1,
prop = 2:5, data = simMVRM, DImodel = "ID",
method = "ML")
print(MVRMmodel)
# Next, we include the simplest interaction structure available in this package, "AV", which adds
# a single extra term per ecosystem function and time point
MVRMmodel_AV <- DImulti(y = 6:8, eco_func = c("Na", "un"), time = c("time", "CS"), unit_IDs = 1,
prop = 2:5, data = simMVRM, DImodel = "AV",
method = "ML")
anova(MVRMmodel, MVRMmodel_AV)
# We select the more model with the lower AIC/BIC value or we use the p-value of the likelihood
# ratio test to determine if we reject the null hypothesis that the extra terms in the model are
# equal to zero, which in this case is lower than our alpha value of 0.05, so we do reject this
# hypothesis and continue with our more complex model.
#
# We can continue increasing the complexity of the interaction structure in the same fashion, this
# time we elect to use the additive interaction structure
MVRMmodel_ADD <- DImulti(y = 6:8, eco_func = c("Na", "un"), time = c("time", "CS"), unit_IDs = 1,
prop = 2:5, data = simMVRM, DImodel = "ADD",
method = "ML")
anova(MVRMmodel_AV, MVRMmodel_ADD)
# We fail to reject the null hypothesis and so we select the average interaction structure.
#
# Finally, we can also increase the model complexity via the inclusion of the non-linear parameter
# theta, which we can estimate, or select a value for. We also choose to estimate using the "REML"
# method as we will do no further fixed effect model comparisons
MVRMmodel_theta <- DImulti(y = 6:8, eco_func = c("Na", "un"), time = c("time", "CS"), unit_IDs = 1,
prop = 2:5, data = simMVRM, DImodel = "AV",
estimate_theta = TRUE, method = "REML")
print(MVRMmodel_theta)
#Finally, we can utilise this model for our interpretation and predictions
head(predict(MVRMmodel_theta))
##################################################################################################
#
##################################################################################################
## Code to simulate data
set.seed(746)
props <- data.frame(plot = integer(),
p1 = integer(),
p2 = integer(),
p3 = integer(),
p4 = integer())
index <- 1 #row number
#Monocultures
for(i in 1:4) #6 species
{
for(j in 1:3) #2 technical reps
{
props[index, i+1] <- 1
index <- index + 1
}
}
#Equal Mixtures
for(rich in sort(rep(2:4, 4))) #3 reps at each richness level
{
sp <- sample(1:4, rich) #randomly pick species from pool
for(j in 1:2) #2 technical reps
{
for(i in sp)
{
props[index, i+1] <- 1/rich #equal proportions
}
index <- index + 1
}
}
#Unequal Mixtures
for(rich in sort(rep(c(2, 3, 4), 50))) #15 reps at each richness level
{
sp <- sample(1:4, rich, replace = TRUE) #randomly pick species from pool
for(j in 1:2) #2 technical reps
{
for(i in 1:4)
{
props[index, i+1] <- sum(sp==i)/rich #equal proportions
}
index <- index + 1
}
}
props[is.na(props)] <- 0
mySimData <- props
ADDs <- DImodels::DI_data(prop=2:5, what=c("ADD"), data=mySimData)
mySimData <- cbind(mySimData, ADDs)
E_AV <- DImodels::DI_data(prop=2:5, what=c("E", "AV"), data=mySimData)
mySimData <- cbind(mySimData, E_AV)
mySimData$plot <- 1:nrow(mySimData)
mySimData$Y1 <- NA
mySimData$Y2 <- NA
mySimData$Y3 <- NA
mySimData$time <- 1
mySimDataT1 <- mySimData
mySimDataT2 <- mySimData
mySimDataT2$time <- 2
nT <- 2 #Number of s
#Principal Components (make sure it's positive definite)
pT <- qr.Q(qr(matrix(stats::rnorm(nT^2), nT)))
ST <- crossprod(pT, pT*(nT:1)) #Sigma
mT <- stats::runif(nT, -0.25, 1.5)
nY <- 3 #Number of Ys
#Principal Components (make sure it's positive definite)
pY <- qr.Q(qr(matrix(stats::rnorm(nY^2), nY)))
SY <- crossprod(pY, pY*(nY:1)) #Sigma
mY <- stats::runif(nY, -0.25, 1.5)
#runif(7, -1, 7) #decide on betas randomly
for(i in 1:nrow(mySimData))
{
#Within subject error
errorT <- MASS::mvrnorm(n=1, mu=mT, Sigma=ST)
errorY <- MASS::mvrnorm(n=1, mu=mY, Sigma=SY)
mySimDataT1$Y1[i] <- -1.0*mySimDataT1$p1[i] + 5.0*mySimDataT1$p2[i] + 2.8*mySimDataT1$p3[i] +
-0.9*mySimDataT1$p4[i] + -0.1*mySimDataT1$E[i] + errorT[1]*errorY[1]
mySimDataT2$Y1[i] <- 0.5*mySimDataT2$p1[i] + 5.4*mySimDataT2$p2[i] + 4.9*mySimDataT2$p3[i] +
-2.1*mySimDataT2$p4[i] + 12.0*mySimDataT1$E[i] + errorT[2]*errorY[1]
mySimDataT1$Y2[i] <- 0.1*mySimDataT1$p1[i] + 4.1*mySimDataT1$p2[i] + -0.5*mySimDataT1$p3[i] +
0.3*mySimDataT1$p4[i] + 2.3*mySimDataT1$E[i] + errorT[1]*errorY[2]
mySimDataT2$Y2[i] <- 2.3*mySimDataT2$p1[i] + 3.2*mySimDataT2$p2[i] + -3.1*mySimDataT2$p3[i] +
2.1*mySimDataT2$p4[i] + 1.6*mySimDataT2$E[i] + errorT[2]*errorY[2]
mySimDataT1$Y3[i] <- 0.9*mySimDataT1$p1[i] + 6.6*mySimDataT1$p2[i] + 3.5*mySimDataT1$p3[i] +
6.1*mySimDataT1$p4[i] + 2.1*mySimDataT1$E[i] + errorT[1]*errorY[3]
mySimDataT2$Y3[i] <- -0.1*mySimDataT2$p1[i] + 7.0*mySimDataT2$p2[i] + 2.8*mySimDataT2$p3[i] +
4.0*mySimDataT2$p4[i] + 6.8*mySimDataT2$E[i] + errorT[2]*errorY[3]
}
mySimData <- rbind(mySimDataT1, mySimDataT2)
mySimData$time <- as.factor(mySimData$time)
mySimData$plot <- as.factor(mySimData$plot)
DImodelsMulti documentation built on May 29, 2024, 2:15 a.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
| simMVRM | R Documentation |
The simulated multivariate repeated measures "simMVRM" dataset
Description
The simMVRM dataset was simulated from a multivariate repeated measures DI model. It contains
336 plots comprising of four species that vary in proportions (p1 - p4). There are
three simulated responses (Y1, Y2, Y3), taken at two differing time points, recorded in a
wide data format (one column per response type). The data was simulated assuming that there were
existing covariances between the responses and between the time pointsand both species identity and
species interaction effects were present.
Usage
data("simMVRM")Format
A data frame with 672 observations on the following 9 variables.
plota factor vector indicating the ID of the experimental unit
p1a numeric vector indicating the initial proportion of species 1
p2a numeric vector indicating the initial proportion of species 2
p3a numeric vector indicating the initial proportion of species 3
p4a numeric vector indicating the initial proportion of species 4
Y1a numeric vector indicating the response of ecosystem function 1
Y2a numeric vector indicating the response of ecosystem function 2
Y3a numeric vector indicating the response of ecosystem function 3
timea factor with levels
12
Details
What are Diversity-Interactions (DI) models?
Diversity-Interactions (DI) models (Kirwan et al., 2009) are a set of tools for analysing and
interpreting data from experiments that explore the effects of species diversity on community-level
responses. We strongly recommend that users read the short introduction to Diversity-Interactions
models (available at: DImodels). Further information on
Diversity-Interactions models is also available in Kirwan et al., 2009 and
Connolly et al., 2013.
Parameter values for the simulation
Multivariate repeated measures DI models take the general form of:
{y}_{kmt} = {Identities}_{kmt} + {Interactions}_{kmt} + {Structures}_{kt} + {\epsilon}_{kmt}
where y are the community-level responses, the Identities are the effects of species
identities for each response and enter the model as individual species proportions measured at the
beginning of the time period, the Interactions are the interactions among the species
proportions, while Structures include other experimental structures such as blocks,
treatments, or density.
The dataset simRM was simulated with:
identity effects for the four species for each time and ecosystem function:
Y1time1 = -1.0, 5.0, 2.8, -0.9
Y1time2 = 0.5, 5.4, 4.9, -2.1
Y2time1 = 0.1, 4.1, -0.5, 0.3
Y2time2 = 2.3, 3.2, -3.1, 2.1
Y3time1 = 0.9, 6.6, 3.5, 6.1
Y3time2 = -0.1, 7.0, 2.8, 4.0
evenness interaction effect for each time and ecosystem function:
Y1time1 = -0.1
Y1time2 = 12.0
Y2time1 = 2.3
Y2time2 = 1.6
Y3time1 = 2.1
Y3time2 = 6.8
-
\epsilonassumed to have a multivaraite normal distribution with mean 0. An ecosystem function matrix Sigma:[1,] 2.36 0.71 -0.29 [2,] 0.71 1.42 -0.39 [3,] -0.29 -0.39 2.21 and a time matrix Sigma:
[1,] 1.69 0.46 [2,] 0.46 1.31
References
Dooley, A., Isbell, F., Kirwan, L., Connolly, J., Finn, J.A. and Brophy, C., 2015.
Testing the effects of diversity on ecosystem multifunctionality using a multivariate model.
Ecology Letters, 18(11), pp.1242-1251.
Finn, J.A., Kirwan, L., Connolly, J., Sebastia, M.T., Helgadottir, A., Baadshaug, O.H.,
Belanger, G., Black, A., Brophy, C., Collins, R.P. and Cop, J., 2013.
Ecosystem function enhanced by combining four functional types of plant species in intensively
managed grassland mixtures: a 3-year continental-scale field experiment.
Journal of Applied Ecology, 50(2), pp.365-375 .
Kirwan, L., Connolly, J., Finn, J.A., Brophy, C., Luscher, A., Nyfeler, D. and Sebastia, M.T.,
2009.
Diversity-interaction modeling: estimating contributions of species identities and interactions
to ecosystem function.
Ecology, 90(8), pp.2032-2038.
Examples
###################################################################################################
###################################################################################################
## Modelling Example
# For a more thorough example of the workflow of this package, please see vignette
# DImulti_workflow using the following code:
# vignette("DImulti_workflow")
head(simMVRM)
# We call DImulti() to fit a series of models, with increasing complexity, and test whether the
# additional terms are worth keeping.
# We begin with an ID DI model, ensuring to use method = "ML" as we will be comparing fixed effects
MVRMmodel <- DImulti(y = 6:8, eco_func = c("Na", "un"), time = c("time", "CS"), unit_IDs = 1,
prop = 2:5, data = simMVRM, DImodel = "ID",
method = "ML")
print(MVRMmodel)
# Next, we include the simplest interaction structure available in this package, "AV", which adds
# a single extra term per ecosystem function and time point
MVRMmodel_AV <- DImulti(y = 6:8, eco_func = c("Na", "un"), time = c("time", "CS"), unit_IDs = 1,
prop = 2:5, data = simMVRM, DImodel = "AV",
method = "ML")
anova(MVRMmodel, MVRMmodel_AV)
# We select the more model with the lower AIC/BIC value or we use the p-value of the likelihood
# ratio test to determine if we reject the null hypothesis that the extra terms in the model are
# equal to zero, which in this case is lower than our alpha value of 0.05, so we do reject this
# hypothesis and continue with our more complex model.
#
# We can continue increasing the complexity of the interaction structure in the same fashion, this
# time we elect to use the additive interaction structure
MVRMmodel_ADD <- DImulti(y = 6:8, eco_func = c("Na", "un"), time = c("time", "CS"), unit_IDs = 1,
prop = 2:5, data = simMVRM, DImodel = "ADD",
method = "ML")
anova(MVRMmodel_AV, MVRMmodel_ADD)
# We fail to reject the null hypothesis and so we select the average interaction structure.
#
# Finally, we can also increase the model complexity via the inclusion of the non-linear parameter
# theta, which we can estimate, or select a value for. We also choose to estimate using the "REML"
# method as we will do no further fixed effect model comparisons
MVRMmodel_theta <- DImulti(y = 6:8, eco_func = c("Na", "un"), time = c("time", "CS"), unit_IDs = 1,
prop = 2:5, data = simMVRM, DImodel = "AV",
estimate_theta = TRUE, method = "REML")
print(MVRMmodel_theta)
#Finally, we can utilise this model for our interpretation and predictions
head(predict(MVRMmodel_theta))
##################################################################################################
#
##################################################################################################
## Code to simulate data
set.seed(746)
props <- data.frame(plot = integer(),
p1 = integer(),
p2 = integer(),
p3 = integer(),
p4 = integer())
index <- 1 #row number
#Monocultures
for(i in 1:4) #6 species
{
for(j in 1:3) #2 technical reps
{
props[index, i+1] <- 1
index <- index + 1
}
}
#Equal Mixtures
for(rich in sort(rep(2:4, 4))) #3 reps at each richness level
{
sp <- sample(1:4, rich) #randomly pick species from pool
for(j in 1:2) #2 technical reps
{
for(i in sp)
{
props[index, i+1] <- 1/rich #equal proportions
}
index <- index + 1
}
}
#Unequal Mixtures
for(rich in sort(rep(c(2, 3, 4), 50))) #15 reps at each richness level
{
sp <- sample(1:4, rich, replace = TRUE) #randomly pick species from pool
for(j in 1:2) #2 technical reps
{
for(i in 1:4)
{
props[index, i+1] <- sum(sp==i)/rich #equal proportions
}
index <- index + 1
}
}
props[is.na(props)] <- 0
mySimData <- props
ADDs <- DImodels::DI_data(prop=2:5, what=c("ADD"), data=mySimData)
mySimData <- cbind(mySimData, ADDs)
E_AV <- DImodels::DI_data(prop=2:5, what=c("E", "AV"), data=mySimData)
mySimData <- cbind(mySimData, E_AV)
mySimData$plot <- 1:nrow(mySimData)
mySimData$Y1 <- NA
mySimData$Y2 <- NA
mySimData$Y3 <- NA
mySimData$time <- 1
mySimDataT1 <- mySimData
mySimDataT2 <- mySimData
mySimDataT2$time <- 2
nT <- 2 #Number of s
#Principal Components (make sure it's positive definite)
pT <- qr.Q(qr(matrix(stats::rnorm(nT^2), nT)))
ST <- crossprod(pT, pT*(nT:1)) #Sigma
mT <- stats::runif(nT, -0.25, 1.5)
nY <- 3 #Number of Ys
#Principal Components (make sure it's positive definite)
pY <- qr.Q(qr(matrix(stats::rnorm(nY^2), nY)))
SY <- crossprod(pY, pY*(nY:1)) #Sigma
mY <- stats::runif(nY, -0.25, 1.5)
#runif(7, -1, 7) #decide on betas randomly
for(i in 1:nrow(mySimData))
{
#Within subject error
errorT <- MASS::mvrnorm(n=1, mu=mT, Sigma=ST)
errorY <- MASS::mvrnorm(n=1, mu=mY, Sigma=SY)
mySimDataT1$Y1[i] <- -1.0*mySimDataT1$p1[i] + 5.0*mySimDataT1$p2[i] + 2.8*mySimDataT1$p3[i] +
-0.9*mySimDataT1$p4[i] + -0.1*mySimDataT1$E[i] + errorT[1]*errorY[1]
mySimDataT2$Y1[i] <- 0.5*mySimDataT2$p1[i] + 5.4*mySimDataT2$p2[i] + 4.9*mySimDataT2$p3[i] +
-2.1*mySimDataT2$p4[i] + 12.0*mySimDataT1$E[i] + errorT[2]*errorY[1]
mySimDataT1$Y2[i] <- 0.1*mySimDataT1$p1[i] + 4.1*mySimDataT1$p2[i] + -0.5*mySimDataT1$p3[i] +
0.3*mySimDataT1$p4[i] + 2.3*mySimDataT1$E[i] + errorT[1]*errorY[2]
mySimDataT2$Y2[i] <- 2.3*mySimDataT2$p1[i] + 3.2*mySimDataT2$p2[i] + -3.1*mySimDataT2$p3[i] +
2.1*mySimDataT2$p4[i] + 1.6*mySimDataT2$E[i] + errorT[2]*errorY[2]
mySimDataT1$Y3[i] <- 0.9*mySimDataT1$p1[i] + 6.6*mySimDataT1$p2[i] + 3.5*mySimDataT1$p3[i] +
6.1*mySimDataT1$p4[i] + 2.1*mySimDataT1$E[i] + errorT[1]*errorY[3]
mySimDataT2$Y3[i] <- -0.1*mySimDataT2$p1[i] + 7.0*mySimDataT2$p2[i] + 2.8*mySimDataT2$p3[i] +
4.0*mySimDataT2$p4[i] + 6.8*mySimDataT2$E[i] + errorT[2]*errorY[3]
}
mySimData <- rbind(mySimDataT1, mySimDataT2)
mySimData$time <- as.factor(mySimData$time)
mySimData$plot <- as.factor(mySimData$plot)
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.