simulateNewData: Simulate New Datasets from MCMC iterations

In ClementCalenge/roedeer3sites: Habitat Selection by the Roe Deer in three French Study Sites

Description

This function allows, for each MCMC iteration, to simulate a hypothetical replication of the dataset using the model; in other words, for each simulated vector of parameters, it simulates a number of relocations in each habitat type for each monitored animal. Such simulated values can be used to compare the observed number of relocations for each roe deer with the statistical distribution of simulated numbers of relocations. THIS FUNCTION CAN ONLY BE USED WITH THE RESULTS OF THE FIT OF MODEL f2 !!!!! i.e. using the dataset coefficientsModel2.

Usage

simulateNewData(coefMatrix, availMatrix, Nlocs, site)

Arguments

coefMatrix	the matrix containing the simulated vectors of parameters, concatenated by rows (i.e. each row is a simulated vector of parameter).
availMatrix	A matrix containing the proportion of the 3 habitat types (columns) for each monitored animal (row)
Nlocs	A vector containing the number of relocations for each animal.
site	A vector containing the site corresponding to each animal.

Value

The result is a list with B elements (B being the number of MCMC iterations, i.e. the number of rows in coefMatrix), each element being a matrix with 3 columns (the 3 habitat types) and N rows (the N animals), containing the simulated number of relocations.

Author(s)

William Gaudry, Clement Calenge, Sonia Said, Jean-Michel Gaillard

Examples

######################################################
##
## 1. Load the data and fit the model

## We load the data used for the fit:
data(HS3sites)

## We remove the information concerning the meadows
## (negligible, see the paper)
HS3sites$locs$meadows <- NULL
HS3sites$hr$meadows <- NULL

## Calculates the total number of relocations
HS3sites$N <- apply(HS3sites$locs,1,sum)

## stores the number of animals
HS3sites$J <- nrow(HS3sites$locs)

## keeps the home-range information in another object
availhr <- HS3sites$hr

## For a better mixing, we scale the covariates
HS3sites$hr <- scale(HS3sites$hr)

## We have described in the example section of the help
## page of the dataset coefficientsModel2 how to fit the model.
## We just load the dataset:
data(coefficientsModel2)


######################################################
##
## 2. Simulate new datasets


## binds the three MCMC chains by row
cor2 <- do.call(rbind, coefficientsModel2)

## Not run: 

## Simulate new datasets: warning! very slow!!
nd <- simulateNewData(cor2, HS3sites$hr, HS3sites$N, HS3sites$site)

## For every number of observed relocations
## identify whether it falls in the 95
## WARNING: can also be very slow!
deds <- matrix(0, nrow=nrow(HS3sites$hr), ncol=3)
for (i in 1:nrow(deds)) {
    for (j in 1:ncol(deds)) {

        ## vector of simulated number of relocations
        vec <- sapply(nd, function(x) x[i,j])

        ## The 95
        qu <- quantile(vec, c(0.025,0.975))

        ## does the observed number fall in the CI?
        deds[i,j] <- as.numeric((HS3sites$locs[i,j]>=qu[1]&HS3sites$locs[i,j]<=qu[2]))
    }
}

## Finally, deds contains 1 if the corresponding number of relocations
## is in the CI. The proportion of CI containing the true value is:
sum(deds)/length(as.vector(deds))




## End(Not run)

ClementCalenge/roedeer3sites documentation built on May 16, 2019, 6:58 p.m.