#####  FOGGING ###############################

#####  TESTS FOR STRUCTURE WITHIN TREES ######

rm(list=ls(all=TRUE))

setwd("C:\\Users\\Tom\\Dropbox\\Research\\Submitted manuscripts\\Ant mosaics in the high canopy of tropical rain forest (Kalsum)\\PeerJ submission\\Data and code for uploading") #change the default folder
require(vegan) #you need to have the vegan package for this to work
data<-read.table(file="Fogging data.txt",header=T) #looks for your file
data #shows your data
as.factor(data$Tree)->data$Tree
as.factor(data$Pan)->data$Pan
n.trees<-nlevels(data$Tree) #number of trees
n.pans<-nlevels(data$Pan) #pans per tree
nsp<-length(data[1,])-2 #number of species
cu.obs<-0 #this will work out the number of checkerboard units in the observed data
for (i in 1:n.trees){ #across all of the trees - do the following...
	nestedchecker(data[(i*n.pans-n.pans+1):(i*n.pans),1:nsp+2])$statistic+cu.obs->cu.obs #take the pans within a tree and works out the number of checkerboard units (CUs) within each tree then adds them all up
	}
cu.obs # the total number of CUs in the observed data within trees
rand<-data #creates another matrix the same as data and names it "rand", this will recieve the randomised data matrixes
nrep=1000 # the number of randomisations that you want to use - the higher this is the longer it will take to run, but the more precise the p-value will be
null.within<-vector(length=nrep) # creates a vector to receive the CU counts for the simulated (randomised) matrices
for (j in 1:nrep){ # for every replicate
	cu.sim<-0 # take cu.sim back to zero
	for (i in 1:n.trees){ # for every tree within every replicate
		commsimulator(data[(i*n.pans-n.pans+1):(i*n.pans),1:nsp+2],method="quasiswap")->rand[(i*n.pans-n.pans+1):(i*n.pans),1:nsp+2] #randomise the data matrix for that tree while keeping the number of species in each pan the same, and the number of occurrences of each species also the same
		nestedchecker(rand[(i*n.pans-n.pans+1):(i*n.pans),1:nsp+2])$statistic+cu.sim->cu.sim #work out the number of CUs for that tree and then go through and add them all up
		}
	cu.sim->null.within[j] # for that replication of the simulation, put the number of CUs observed into null.within
	}
#hist(null.within) #draw a histogram of null.within
#abline (v=cu.obs) #adds a vertical line where your observed number of CUs is
write.table(null.within,file="Mosaics fogging within.out.txt") #writes the simulated CU counts to a text file at this location - you can use this data file to draw a histogram of your results
p.upper.within<-length(subset(null.within,null.within>cu.obs))/nrep #works out the p-value for there being significantly more CUs than you would expect at random
p.lower.within<-length(subset(null.within,null.within<cu.obs))/nrep #the same for there being fewer CUs
p.upper.within #list of upper tail p-values
p.lower.within #list of lower tail p-values

###### TESTS FOR STRUCTURE BETWEEN TREES #######

#This does exactly as above, but concoct randomised trees first using sets of 10 pans, none of which are from the same tree

n.matr<-10
p.upper.between<-vector(length=n.matr)
p.lower.between<-vector(length=n.matr)
for (k in 1:n.matr){
	diff.tree.pans<-data
	for (i in 1:n.trees){
		sample(c(1:n.trees),n.pans,replace=FALSE)->selected.trees
		data[selected.trees,]->diff.tree.pans[(i*n.pans-n.pans+1):(i*n.pans),]
		}
	diff.tree.pans
	cu.between.obs<-0 # the code from here down is just copied from above with some minor changes - it tests the within "tree" cu counts for significance, so depending on how many constructed datasets you use you will end up with that many p-values
	for (i in 1:n.trees){
		nestedchecker(diff.tree.pans[(i*n.pans-n.pans+1):(i*n.pans),1:nsp+2])$statistic+cu.between.obs->cu.between.obs
		}
	rand.between<-data
	null.between<-vector(length=nrep)
	for (j in 1:nrep){
		cu.between.sim<-0
		for (i in 1:n.trees){
			commsimulator(diff.tree.pans[(i*n.pans-n.pans+1):(i*n.pans),1:nsp+2],method="quasiswap")->rand[(i*n.pans-n.pans+1):(i*n.pans),1:nsp+2]
			nestedchecker(rand[(i*n.pans-n.pans+1):(i*n.pans),1:nsp+2])$statistic+cu.between.sim->cu.between.sim
			}
		cu.between.sim->null.between[j]
		}
	p.upper.between[k]<-length(subset(null.between,null.between>cu.between.obs))/nrep
	p.lower.between[k]<-length(subset(null.between,null.between<cu.between.obs))/nrep
	}
cu.between.obs
#hist(null.between) #draw a histogram of null.between
#abline (v=cu.between.obs) #adds a vertical line where your observed number of CUs is
write.table(null.between,file="Mosaics fogging between.out.txt") #writes the simulated CU counts to a text file at this location - you can use this data file to draw a histogram of your results. Note that this is just one example histogram (the last) out of all ten replicates generated
p.upper.between #list of upper tail p-values
p.lower.between #list of lower tail p-values

############## BAITING #####################

######### NIGHT ##########



#####  TESTS FOR STRUCTURE WITHIN TREES ######



rm(list=ls(all=TRUE))
setwd("C:\\Users\\Tom\\Dropbox\\Research\\Submitted manuscripts\\Ant mosaics in the high canopy of tropical rain forest (Kalsum)\\PeerJ submission\\Data and code for uploading") #change the default folder

require(vegan) #you need to have the vegan package for this to work


data<-read.table(file="Baiting data (night).txt",header=T) #load data

#data #shows your data

as.factor(data$Tree)->data$Tree

as.factor(data$Pan)->data$Pan

n.trees<-nlevels(data$Tree) #number of trees

n.pans<-nlevels(data$Pan) #pans per tree

nsp<-length(data[1,])-2 #number of species

data[,3:nsp+2][data[,3:nsp+2]>0]<-1

cu.obs<-0 #this will work out the number of checkerboard units in the observed data

for (i in 1:n.trees){ #across all of the trees - do the following...

	nestedchecker(data[(i*n.pans-n.pans+1):(i*n.pans),1:nsp+2])$statistic+cu.obs->cu.obs #take the pans within a tree and works out the number of checkerboard units (CUs) within each tree then adds them all up

	}

cu.obs # the total number of CUs in the observed data within trees

rand<-data #creates another matrix the same as data and names it "rand", this will recieve the randomised data matrixes

nrep=1000 # the number of randomisations that you want to use - the higher this is the longer it will take to run, but the more precise the p-value will be

null.within<-vector(length=nrep) # creates a vector to receive the CU counts for the simulated (randomised) matrices

for (j in 1:nrep){ # for every replicate

	cu.sim<-0 # take cu.sim back to zero

	for (i in 1:n.trees){ # for every tree within every replicate

		commsimulator(data[(i*n.pans-n.pans+1):(i*n.pans),1:nsp+2],method="quasiswap")->rand[(i*n.pans-n.pans+1):(i*n.pans),1:nsp+2] #randomise the data matrix for that tree while keeping the number of species in each pan the same, and the number of occurrences of each species also the same

		nestedchecker(rand[(i*n.pans-n.pans+1):(i*n.pans),1:nsp+2])$statistic+cu.sim->cu.sim #work out the number of CUs for that tree and then go through and add them all up

		}

	cu.sim->null.within[j] # for that replication of the simulation, put the number of CUs observed into null.within

	}
cu.obs

#hist(null.within) #draw a histogram of null.within

#abline (v=cu.obs) #adds a vertical line where your observed number of CUs is

write.table(null.within,file="Mosaics baiting night within.out.txt") #writes the simulated CU counts to a text file at this location - you can use this data file to draw a histogram of your results

p.upper.within<-length(subset(null.within,null.within>cu.obs))/nrep #works out the p-value for there being significantly more CUs than you would expect at random

p.lower.within<-length(subset(null.within,null.within<cu.obs))/nrep #the same for there being fewer CUs

p.upper.within

p.lower.within



###### TESTS FOR STRUCTURE BETWEEN TREES #######



#This does exactly as above, but concoct randomised trees first using sets of 10 pans, none of which are from the same tree

n.matr<-10

p.upper.selected<-vector(length=n.matr)

p.lower.selected<-vector(length=n.matr)

for (k in 1:n.matr){

	diff.tree.pans<-data

	for (i in 1:n.trees){

		sample(c(1:n.trees),n.pans,replace=FALSE)->selected.trees

		data[selected.trees,]->diff.tree.pans[(i*n.pans-n.pans+1):(i*n.pans),]

		}

	diff.tree.pans

	cu.between.obs<-0 # the code from here down is just copied from above with some minor changes - it tests the within "tree" cu counts for significance, so depending on how many constructed datasets you use you will end up with that many p-values

	for (i in 1:n.trees){

		nestedchecker(diff.tree.pans[(i*n.pans-n.pans+1):(i*n.pans),1:nsp+2])$statistic+cu.between.obs->cu.between.obs

		}

	rand.between<-data

	null.between<-vector(length=nrep)

	for (j in 1:nrep){

		cu.between.sim<-0

		for (i in 1:n.trees){

			commsimulator(diff.tree.pans[(i*n.pans-n.pans+1):(i*n.pans),1:nsp+2],method="quasiswap")->rand[(i*n.pans-n.pans+1):(i*n.pans),1:nsp+2]

			nestedchecker(rand[(i*n.pans-n.pans+1):(i*n.pans),1:nsp+2])$statistic+cu.between.sim->cu.between.sim

			}

		cu.between.sim->null.between[j]

		}

	p.upper.selected[k]<-length(subset(null.between,null.between>cu.between.obs))/nrep

	p.lower.selected[k]<-length(subset(null.between,null.between<cu.between.obs))/nrep

	}
cu.between.obs
#hist(null.between) #draw a histogram of null.between

#abline (v=cu.between.obs) #adds a vertical line where your observed number of CUs is

p.upper.selected #list of upper tail p-values

p.lower.selected #list of lower tail p-values

write.table(null.between,file="Mosaics baiting night between.out.txt") #writes the simulated CU counts to a text file at this location - you can use this data file to draw a histogram of your results


######### DAY ##########



#####  TESTS FOR STRUCTURE WITHIN TREES ######



rm(list=ls(all=TRUE))
setwd("C:\\Users\\Tom\\Dropbox\\Research\\Submitted manuscripts\\Ant mosaics in the high canopy of tropical rain forest (Kalsum)\\PeerJ submission\\Data and code for uploading") #change the default folder

require(vegan) #you need to have the vegan package for this to work


data<-read.table(file="Baiting data (day).txt",header=T) #use this line of code for the analysis of the day ants, otherwise everything else is the same, apart from writing the correct output file

#data #shows your data

as.factor(data$Tree)->data$Tree

as.factor(data$Pan)->data$Pan

n.trees<-nlevels(data$Tree) #number of trees

n.pans<-nlevels(data$Pan) #pans per tree

nsp<-length(data[1,])-2 #number of species

data[,3:nsp+2][data[,3:nsp+2]>0]<-1

cu.obs<-0 #this will work out the number of checkerboard units in the observed data

for (i in 1:n.trees){ #across all of the trees - do the following...

	nestedchecker(data[(i*n.pans-n.pans+1):(i*n.pans),1:nsp+2])$statistic+cu.obs->cu.obs #take the pans within a tree and works out the number of checkerboard units (CUs) within each tree then adds them all up

	}

cu.obs # the total number of CUs in the observed data within trees

rand<-data #creates another matrix the same as data and names it "rand", this will recieve the randomised data matrixes

nrep=1000 # the number of randomisations that you want to use - the higher this is the longer it will take to run, but the more precise the p-value will be

null.within<-vector(length=nrep) # creates a vector to receive the CU counts for the simulated (randomised) matrices

for (j in 1:nrep){ # for every replicate

	cu.sim<-0 # take cu.sim back to zero

	for (i in 1:n.trees){ # for every tree within every replicate

		commsimulator(data[(i*n.pans-n.pans+1):(i*n.pans),1:nsp+2],method="quasiswap")->rand[(i*n.pans-n.pans+1):(i*n.pans),1:nsp+2] #randomise the data matrix for that tree while keeping the number of species in each pan the same, and the number of occurrences of each species also the same

		nestedchecker(rand[(i*n.pans-n.pans+1):(i*n.pans),1:nsp+2])$statistic+cu.sim->cu.sim #work out the number of CUs for that tree and then go through and add them all up

		}

	cu.sim->null.within[j] # for that replication of the simulation, put the number of CUs observed into null.within

	}

#hist(null.within) #draw a histogram of null.within

#abline (v=cu.obs) #adds a vertical line where your observed number of CUs is

write.table(null.within,file="Mosaics baiting day within.out.txt") #writes the simulated CU counts to a text file at this location - you can use this data file to draw a histogram of your results

p.upper.within<-length(subset(null.within,null.within>cu.obs))/nrep #works out the p-value for there being significantly more CUs than you would expect at random

p.lower.within<-length(subset(null.within,null.within<cu.obs))/nrep #the same for there being fewer CUs

p.upper.within

p.lower.within



###### TESTS FOR STRUCTURE BETWEEN TREES #######



#This does exactly as above, but concoct randomised trees first using sets of 10 pans, none of which are from the same tree

n.matr<-10

p.upper.selected<-vector(length=n.matr)

p.lower.selected<-vector(length=n.matr)

for (k in 1:n.matr){

	diff.tree.pans<-data

	for (i in 1:n.trees){

		sample(c(1:n.trees),n.pans,replace=FALSE)->selected.trees

		data[selected.trees,]->diff.tree.pans[(i*n.pans-n.pans+1):(i*n.pans),]

		}

	diff.tree.pans

	cu.between.obs<-0 # the code from here down is just copied from above with some minor changes - it tests the within "tree" cu counts for significance, so depending on how many constructed datasets you use you will end up with that many p-values

	for (i in 1:n.trees){

		nestedchecker(diff.tree.pans[(i*n.pans-n.pans+1):(i*n.pans),1:nsp+2])$statistic+cu.between.obs->cu.between.obs

		}

	rand.between<-data

	null.between<-vector(length=nrep)

	for (j in 1:nrep){

		cu.between.sim<-0

		for (i in 1:n.trees){

			commsimulator(diff.tree.pans[(i*n.pans-n.pans+1):(i*n.pans),1:nsp+2],method="quasiswap")->rand[(i*n.pans-n.pans+1):(i*n.pans),1:nsp+2]

			nestedchecker(rand[(i*n.pans-n.pans+1):(i*n.pans),1:nsp+2])$statistic+cu.between.sim->cu.between.sim

			}

		cu.between.sim->null.between[j]

		}

	p.upper.selected[k]<-length(subset(null.between,null.between>cu.between.obs))/nrep

	p.lower.selected[k]<-length(subset(null.between,null.between<cu.between.obs))/nrep

	}
cu.between.obs
#hist(null.between) #draw a histogram of null.between

#abline (v=cu.between.obs) #adds a vertical line where your observed number of CUs is

p.upper.selected #list of upper tail p-values

p.lower.selected #list of lower tail p-values

write.table(null.between,file="Mosaics baiting day between.out.txt") #writes the simulated CU counts to a text file at this location - you can use this data file to draw a histogram of your results

############PLOTTING

par(mfrow=c(3,2))



data<-read.table(file="Mosaics baiting day within.out.txt",header=T)

hist(data$x,main=NA,xlab="Checkerboard units",xlim=c(720,820),breaks=c((0:20)*5+720),col="light grey")

abline(v=767,lty=9,lwd=5)





data<-read.table(file="Mosaics baiting day between.out.txt",header=T)

hist(data$x,main=NA,xlab="Checkerboard units",xlim=c(560,660),breaks=c((0:20)*5+560),col="light grey")

abline(v=657,lty=9,lwd=5)



data<-read.table(file="Mosaics baiting night within.out.txt",header=T)

hist(data$x,main=NA,xlab="Checkerboard units", xlim=c(730,830),breaks=c((0:20)*5+730),col="light grey")

abline(v=818,lty=9,lwd=5)



data<-read.table(file="Mosaics baiting night between.out.txt",header=T)

hist(data$x,main=NA,xlab="Checkerboard units", xlim=c(1030,1130),breaks=c((0:20)*5+1030),col="light grey")

abline(v=1111,lty=9,lwd=5)




data<-read.table(file="Mosaics fogging within.out.txt",header=T)

hist(data$x,main=NA,xlab="Checkerboard units",xlim=c(9200,9900),breaks=c((0:20)*35+9200),col="light grey")

abline(v=9806,lty=9,lwd=5)



data<-read.table(file="Mosaics fogging between.out.txt",header=T)

hist(data$x,main=NA,xlab="Checkerboard units",xlim=c(10200,11000),breaks=c((0:20)*35+10200),col="light grey")

abline(v=10935,lty=9,lwd=5)