MetaLandSim (version 1.0.4)

simulatedifm: Set of simulated data to use with the IFM parameter estimation functions. The data were generated using the code provided in "details".

Description

This dataset loads several objects:

'sim.area', 'sim.det.20', 'sim.distance', 'z.sim', 'z.sim.20' and 'z.sim.20.fa'.

Format

sim.area

Vector of the areas for each site; here, 100 sites.

sim.det.20

100 x 10 x 3 array corresponding to nsites x nyears x nvisits. Data simulated with year-specific detection probabilities equal to 0.4,0.6,0.2,0.9,0.3,0.4,0.6,0.2,0.9,0.3.

sim.distance

nsite x nsite distance matrix.

z.sim

nyear x nsite occupancy data generated with perfect detection.

z.sim.20

nyear x nsite occupancy data generated with perfect detection with approximately 20% of data missing at random.

z.sim.20.fa

nyear x nsite occupancy data containing false absences, which can be used to explore the bias of ifm.missing.MCMC and ifm.naive.MCMC when there is imperfect detection.

Details

These datasets were created using the code included in Examples.

Examples

# NOT RUN {
# }
# NOT RUN {
##############################################
# Areas for 100 hundred sites were created from a log normal distribution with mean
# and variance equal to the mean and variance of the area of the sites in the
# Sierra Foothills black rail population.
# Universal Transverse Mercator locations (UTMs) for each site were
# simulated from the mean and variance of the UTM Northing and Easting
#
# Dynamics were simulated for 1000 years. The parameters were chosen such that the metapopulation
# persisted with reasonable turnover. The last 10 years of these data were retained.
# For the detection data sets, we simulated a removal design based on three visits.

#-------------------
# IFM SIMULATE
#------------------
set.seed(123)

#-------
#AREAS
#-------
mean.log.area=-0.75
sd.log.area=1.33
nsite=100
log.sim.area=rnorm(nsite,mean.log.area,sd.log.area)
sim.area = exp(log.sim.area)

#---------
#DISTANCE
#--------
sim.site.x=rnorm(nsite,643930,9000)
sim.site.y=rnorm(nsite,4340500,10500)

sim.site.x.mat.col=matrix(rep(sim.site.x,nsite),ncol=nsite,byrow=TRUE)
sim.site.x.mat.row=matrix(rep(sim.site.x,nsite),ncol=nsite)
sim.site.delta.x=(sim.site.x.mat.col-sim.site.x.mat.row)^2

sim.site.y.mat.col=matrix(rep(sim.site.y,nsite),ncol=nsite,byrow=TRUE)
sim.site.y.mat.row=matrix(rep(sim.site.y,nsite),ncol=nsite)
sim.site.delta.y=(sim.site.y.mat.col-sim.site.y.mat.row)^2

# scale distance so that alpha = 2 is reasonable:
sim.distance=((sim.site.delta.x+sim.site.delta.y)^0.5)/100000
diag(sim.distance)=99

#-------------#
# CREATE SPOM #
#-------------#
spom=function(nsite,nyear.sim,alpha,b,y,e,x) {
	psi=matrix(rep(NA,nsite*nyear.sim),ncol=nyear.sim)
	psi1=rbinom(nsite,1,0.8)
	psi[,1]=psi1
	s.i.temp=exp(-alpha*sim.distance)
	s.i.temp[s.i.temp==1]=0
	e.i=e/sim.area^x
	e.i[e.i>1]=1

	for (i in 2:nyear.sim) {
		s.i=s.i.temp<!-- %*%((sim.area^b)*psi[,i-1]) -->
		c.i=s.i^2/(s.i^2+y^2)
		e.i.re=e.i*(1-c.i)
		mu1=psi[,i-1]*(1-e.i.re)+(1-psi[,i-1])*c.i
		psi.temp=rbinom(nsite,1,mu1)
		psi[,i]=psi.temp
		}
	psi
	}

#---------------
# SIMULATE IFM
#---------------

nyear.sim=1000
psi.sim=spom(nsite,nyear.sim,alpha=20,b=0.5,y=7.5,e=0.25,x=0.25)

nyear=10

# Data from this dataset conforms to the assumptions of IFM Naive:
z.sim=psi.sim[,(nyear.sim+1-nyear):nyear.sim]
apply(z.sim,2,mean)


#CREATE DETECTION HISTORY
p=rep(c(0.4,0.6,0.2,0.9,0.3),2)
nrep=3
temp=rep(1,nrep*nsite*nyear)
p.mat=matrix(rep(p,nsite),nrow=nsite,byrow=TRUE)
temp.z.sim=z.sim*p.mat
sim.det=rbinom(temp,1,temp.z.sim)
dim(sim.det)=c(nsite,nyear,nrep)

#ENFORCE REMOVAL DESIGN
for (i in 1:nsite) {
  for(t in 1:nyear) {
    if (sim.det[i,t,1]==1) sim.det[i,t,2]=2
    if (sim.det[i,t,1]==1) sim.det[i,t,3]=2
    if (sim.det[i,t,2]==1) sim.det[i,t,3]=2
  }
}

sim.det[sim.det==2]=NA
sim.det.no.missing.values=sim.det

#RANDOMLY CREATE MISSING DATA; 20<!-- % -->
# Data are missing when a site was never visited in a given year
unif.mat=runif(nyear*nsite)
z.sim.20=z.sim
z.sim.20[unif.mat<0.2]=NA

sim.det.20=sim.det
sim.det.20[rep(is.na(z.sim.20),nrep)]=NA

#CREATE DATASET WITH MISSING VALUES AND FALSE ABSENCES
#THIS IS TO CHECK HOW IFM.NAIVE AND IFM.MISSING
# LEAD TO BIASES
z.sim.20.fa = apply(sim.det.20,c(1,2),sum,na.rm=TRUE)
z.sim.20.fa[unif.mat<0.20]=NA

# z.sim: Perfect detection, one visit per year.
# z.sim.20: Perfect detection, but 20<!-- % of sites have missing data. -->
# z.sim.20.fa: 20<!-- % of sites have missing data in each year, sites are visited according to the removal design, -->
#   then collapsed to a single observation per year equal to one if a detection occurred.
# sim.det.20: 20<!-- % of sites have missing data, there is imperfect detection, and the data follow a removal design. -->
#   The data are arranged in a 3-d array sites x years x visits

save(z.sim,z.sim.20,z.sim.20.fa,sim.det.
20,sim.distance,sim.area,file=paste("SIMULATE_DATA_MDL_CMP",nsite,"_",nyear,
"_",nrep,".RData",sep=""))
# }
# NOT RUN {
data(simulatedifm)
ls()

# }