### First read in the data files
lang.d<-read.csv("LanguageDistances.csv",row.names=1,header=T,comment="",quote="\"",stringsAsFactors=F,check.names=F)
space.d<-read.csv("Geographic_Distance.csv",row.names=1,header=T,comment="",quote="\"",stringsAsFactors=F,check.names=F)
space.adj<-read.csv("geo_adjacency_matrix.csv",row.names=1,header=T,comment="",quote="\"",stringsAsFactors=F,check.names=F)
### check names are full intersection and matrices are aligned
nrow(lang.d)
ncol(lang.d)
nrow(space.d)
ncol(space.d)
nrow(space.adj)
ncol(space.adj)
length(intersect(row.names(lang.d),colnames(lang.d)))
length(intersect(row.names(lang.d),row.names(space.d)))
length(intersect(row.names(lang.d),colnames(space.d)))
length(intersect(row.names(lang.d),row.names(space.adj)))
length(intersect(row.names(lang.d),colnames(space.adj)))

space.d<-space.d[row.names(lang.d),colnames(lang.d)]
space.adj<-space.adj[row.names(lang.d),colnames(lang.d)]

lang.adj<-lang.d
lang.adj[lang.d>1]<-0
lang.adj[lang.adj>0]<-1

lang.d.norm<-(max(lang.d)-lang.d)/max(lang.d)
lang.d.norm[is.na(lang.d.norm)]<-0
diag(lang.d.norm)<-0

space.d.norm<-(max(space.d)-space.d)/max(space.d)
space.d.norm[is.na(space.d.norm)]<-0
diag(space.d.norm)<-0

diag(lang.adj)<-0
lang.adj.rnorm<-lang.adj/rowSums(lang.adj)
lang.adj.rnorm[is.na(lang.adj.rnorm)]<-0

diag(space.adj)<-0
space.adj.rnorm<-space.adj/rowSums(space.adj)
space.adj.rnorm[is.na(space.adj.rnorm)]<-0

#source("FunctionSimCharOnNetworkContinuousModified.txt")
#sim.contin.on.network(file.name="lang.d",network=lang.d.norm,influence=0.2,Nchar=1,Ntaxa=nrow(lang.d.norm),Nmat=1000)
#sim.contin.on.network(file.name="space.d",network=space.d.norm,influence=0.2,Nchar=1,Ntaxa=nrow(space.d.norm),Nmat=1000)
#sim.contin.on.network(file.name="lang.adj",network=lang.adj.rnorm,influence=0.2,Nchar=1,Ntaxa=nrow(lang.adj.rnorm),Nmat=1000)
#sim.contin.on.network(file.name="space.adj",network=space.adj.rnorm,influence=0.2,Nchar=1,Ntaxa=nrow(space.adj.rnorm),Nmat=1000)

### Here you pick which sims to load
#load("lang.adj_network.sims.array.continuous")
#load("space.adj_network.sims.array.continuous")
#load("lang.d_network.sims.array.continuous")
#load("space.d_network.sims.array.continuous")



netsims<-allsims

Nmat<-dim(netsims)[3]

netsims<-allsims
dim(netsims)
pc.comps<-matrix(nrow=dim(netsims)[1],ncol=Nmat)

for(i in 1:Nmat) {
pc.comps[,i]<-netsims[,,i]
			}




bic.la.ld.sa.sd<-c()
bic.la.ld.sa<-c()
bic.la.ld.sd<-c()
bic.la.sa.sd<-c()
bic.ld.sa.sd<-c()
bic.la.ld<-c()
bic.la.sa<-c()
bic.ld.sa<-c()
bic.sa.sd<-c()
bic.sd.la<-c()
bic.sd.ld<-c()
bic.la<-c()
bic.ld<-c()
bic.sa<-c()
bic.sd<-c()
bic.null<-c()


library(sna)


int<-rep(1,nrow(pc.comps))


for(i in 1:250) {

net.reg.la.ld.sa.sd<-lnam(pc.comps[,i],int,W2=list(lang.adj.rnorm,lang.d.norm,space.adj.rnorm,space.d.norm))

net.reg.la.ld.sa<-lnam(pc.comps[,i],int,W2=list(lang.adj.rnorm,lang.d.norm,space.adj.rnorm))

net.reg.la.ld.sd<-lnam(pc.comps[,i],int,W2=list(lang.adj.rnorm,lang.d.norm,space.d.norm))

net.reg.la.sa.sd<-lnam(pc.comps[,i],int,W2=list(lang.adj.rnorm,space.adj.rnorm,space.d.norm))

net.reg.ld.sa.sd<-lnam(pc.comps[,i],int,W2=list(lang.d.norm,space.adj.rnorm,space.d.norm))

net.reg.la.ld<-lnam(pc.comps[,i],int,W2=list(lang.adj.rnorm,lang.d.norm))
net.reg.la.sa<-lnam(pc.comps[,i],int,W2=list(lang.adj.rnorm,space.adj.rnorm))
net.reg.ld.sa<-lnam(pc.comps[,i],int,W2=list(lang.d.norm,space.adj.rnorm))
net.reg.sa.sd<-lnam(pc.comps[,i],int,W2=list(space.adj.rnorm,space.d.norm))
net.reg.sd.la<-lnam(pc.comps[,i],int,W2=list(lang.adj.rnorm,space.d.norm))
net.reg.sd.ld<-lnam(pc.comps[,i],int,W2=list(lang.d.norm,space.d.norm))

net.reg.la<-lnam(pc.comps[,i],int,W2=list(lang.adj.rnorm))
net.reg.ld<-lnam(pc.comps[,i],int,W2=list(lang.d.norm))
net.reg.sa<-lnam(pc.comps[,i],int,W2=list(space.adj.rnorm))
net.reg.sd<-lnam(pc.comps[,i],int,W2=list(space.d.norm))

bic.la.ld.sa.sd[i]<-(-2)*net.reg.la.ld.sa.sd$lnlik.model+log(nrow(pc.comps))*net.reg.la.ld.sa.sd$df.model

bic.la.ld.sa[i]<-(-2)*net.reg.la.ld.sa$lnlik.model+log(nrow(pc.comps))*net.reg.la.ld.sa$df.model

bic.la.ld.sd[i]<-(-2)*net.reg.la.ld.sd$lnlik.model+log(nrow(pc.comps))*net.reg.la.ld.sd$df.model

bic.la.sa.sd[i]<-(-2)*net.reg.la.sa.sd$lnlik.model+log(nrow(pc.comps))*net.reg.la.sa.sd$df.model

bic.ld.sa.sd[i]<-(-2)*net.reg.ld.sa.sd$lnlik.model+log(nrow(pc.comps))*net.reg.ld.sa.sd$df.model

bic.la.ld[i]<-(-2)*net.reg.la.ld$lnlik.model+log(nrow(pc.comps))*net.reg.la.ld$df.model

bic.la.sa[i]<-(-2)*net.reg.la.sa$lnlik.model+log(nrow(pc.comps))*net.reg.la.sa$df.model

bic.ld.sa[i]<-(-2)*net.reg.ld.sa$lnlik.model+log(nrow(pc.comps))*net.reg.ld.sa$df.model

bic.sa.sd[i]<-(-2)*net.reg.sa.sd$lnlik.model+log(nrow(pc.comps))*net.reg.sa.sd$df.model

bic.sd.la[i]<-(-2)*net.reg.sd.la$lnlik.model+log(nrow(pc.comps))*net.reg.sd.la$df.model

bic.sd.ld[i]<-(-2)*net.reg.sd.ld$lnlik.model+log(nrow(pc.comps))*net.reg.sd.ld$df.model

bic.la[i]<-(-2)*net.reg.la$lnlik.model+log(nrow(pc.comps))*net.reg.la$df.model

bic.ld[i]<-(-2)*net.reg.ld$lnlik.model+log(nrow(pc.comps))*net.reg.ld$df.model

bic.sa[i]<-(-2)*net.reg.sa$lnlik.model+log(nrow(pc.comps))*net.reg.sa$df.model

bic.sd[i]<-(-2)*net.reg.sd$lnlik.model+log(nrow(pc.comps))*net.reg.sd$df.model


bic.null[i]<-(-2)*net.reg.sd$lnlik.null+log(nrow(pc.comps))*net.reg.sd$df.null

print(i)
}




top.models<-matrix(nrow=length(bic.la.ld.sa.sd),ncol=8)

model.name.list<-c("bic.la.ld.sa.sd","bic.la.ld.sa","bic.la.ld.sd","bic.la.sa.sd","bic.ld.sa.sd","bic.la.ld","bic.la.sa","bic.ld.sa","bic.sa.sd","bic.sd.la","bic.sd.ld","bic.la","bic.ld","bic.sa","bic.sd","bic.null")

for (i in 1:length(bic.la.ld.sa.sd)) {

temp.list<-c(bic.la.ld.sa.sd[i],bic.la.ld.sa[i],bic.la.ld.sd[i],bic.la.sa.sd[i],bic.ld.sa.sd[i],bic.la.ld[i],bic.la.sa[i],bic.ld.sa[i],bic.sa.sd[i],bic.sd.la[i],bic.sd.ld[i],bic.la[i],bic.ld[i],bic.sa[i],bic.sd[i],bic.null[i])

 names(temp.list)<-model.name.list

 top.models[i,1]<-names(temp.list)[which.min(temp.list)]
 top.models[i,5]<-min(temp.list)

 temp.list2<-temp.list[temp.list!=min(temp.list)]

 top.models[i,2]<-names(temp.list2)[which.min(temp.list2)]
 top.models[i,6]<-min(temp.list2)

 temp.list3<-temp.list2[temp.list2!=min(temp.list2)]

 top.models[i,3]<-names(temp.list3)[which.min(temp.list3)]
 top.models[i,7]<-min(temp.list3)


 temp.list4<-temp.list3[temp.list3!=min(temp.list3)]

top.models[i,4]<-names(temp.list4)[which.min(temp.list4)]
 top.models[i,4]<-names(temp.list4)[which.min(temp.list4)]
 top.models[i,8]<-min(temp.list4)

 }

length(table(top.models[,1]))
length(table(top.models[,2]))
table(top.models[,1])
table(top.models[,2])

median(as.numeric(na.omit(top.models[,6]))-as.numeric(na.omit(top.models[,5])))

median(as.numeric(na.omit(top.models[,7]))-as.numeric(na.omit(top.models[,5])))

#write.table(top.models,file="lang.adj.sims.processed.txt",row.names=F,col.names=F,sep="|",quote=F)

#write.table(top.models,file="space.adj.sims.processed.txt",row.names=F,col.names=F,sep="|",quote=F)

#write.table(top.models,file="space.d.sims.processed.txt",row.names=F,col.names=F,sep="|",quote=F)

#write.table(top.models,file="lang.d.sims.processed.txt",row.names=F,col.names=F,sep="|",quote=F)