###############################################################################
## Creat the interaction variables

setwd("D:\\5poultry\\logic\\0CreatCom4") 
d <- read.csv("environment.csv")

##  popdenglwd
zpopden <- (d$popden - mean(d$popden))/sd(d$popden)
nzpopden <-(zpopden-min(zpopden))/(max(zpopden)-min(zpopden))
zglwd <- ((d$glwd - mean(d$glwd))/sd(d$glwd))
nzglwd <-(zglwd-min(zglwd))/(max(zglwd)-min(zglwd))

anzglwd <- (1-nzglwd)

popdenglwd <- nzpopden*anzglwd*100

write.csv(popdenglwd,file="popdenglwd.csv")

##  popdenramsar
zpopden <- (d$popden - mean(d$popden))/sd(d$popden)
nzpopden <-(zpopden-min(zpopden))/(max(zpopden)-min(zpopden))
zramsar <- ((d$ramsar - mean(d$ramsar))/sd(d$ramsar))
nzramsar <-(zramsar-min(zramsar))/(max(zramsar)-min(zramsar))

anzramsar <- (1-nzramsar)

popdenramsar <- nzpopden*anzramsar*100

write.csv(popdenramsar,file="popdenramsar.csv")

##  pouldenglwd

zpoultry <- (d$poultry - mean(d$poultry))/sd(d$poultry)
nzpoultry <-(zpoultry-min(zpoultry))/(max(zpoultry)-min(zpoultry))
zglwd <- ((d$glwd - mean(d$glwd))/sd(d$glwd))
nzglwd <-(zglwd-min(zglwd))/(max(zglwd)-min(zglwd))

anzglwd <- (1-nzglwd)

poultryglwd <- nzpoultry*anzglwd*100

write.csv(popdenglwd,file="poultryglwd.csv")

##  poultryramsar
zpoultry <- (d$poultry - mean(d$poultry))/sd(d$poultry)
nzpoultry <-(zpoultry-min(zpoultry))/(max(zpoultry)-min(zpoultry))
zramsar <- ((d$ramsar - mean(d$ramsar))/sd(d$ramsar))
nzramsar <-(zramsar-min(zramsar))/(max(zramsar)-min(zramsar))

anzramsar <- (1-nzramsar)

poultryramsar <- nzpoultry*anzramsar*100

write.csv(poultryramsar,file="poultryramsar.csv")

## make the file "envrionment_com.csv" which includes 4 interaction variables
###############################################################################
## Univariate analysis ~ linear effect

library(rms)
library(PresenceAbsence)   

## variables: city metropolis mjroad road
## glwd ramsar popden poultry dem aspect slope mapet
## maridity pre01 pre02 pre03 pre04 pre05 pre06 pre07 
## pre08 pre09 pre10 pre11 pre12 tmin01 tmin02 tmin03 tmin04 tmin05 tmin06 tmin07 tmin08 tmin09 
## tmin10 tmin11 tmin12 tmax01 tmax02 tmax03 tmax04 tmax05 tmax06 tmax07
## tmax08 tmax09 tmax10 tmax11 tmax12 NDVI01 NDVI02 NDVI03 NDVI04 NDVI05
## NDVI06 NDVI07 NDVI08 NDVI09 NDVI10 NDVI11 NDVI12 
## popdenglwd  popdenramsar  poultryglwd  poultryramsar  
## change 11 times when use another variable

setwd("D:\\5Poultry\\logic\\1univariate") 

bootnum=1000

d <- read.csv("environment_com.csv")
                                                                        
output<-data.frame(matrix(nrow=bootnum,ncol=13))
names(output)=c('EXP-B_poultryramsar','CIExpB-Low_poultryramsar','CIExpB-Up_poultryramsar','P_poultryramsar','R2','AIC',"AUC","AUC_LCI", "AUC_UCI",'P-Value','Threshold','Kappa','Kappa_sd')

for (i in 1:bootnum){

P<-subset(d,select=c(names(d)),d$pa==1)
A<-subset(d,select=c(names(d)),d$pa==0)              

Arind<-sample(1:nrow(A),245,replace=T)
Prind<-sample(1:nrow(P),245,replace=T)
Ain<-A[Arind,]
Pin<-P[Prind,]
dIN<- rbind(Ain,Pin)

poultryramsar <- (dIN$poultryramsar - mean(d$poultryramsar))/sd(d$poultryramsar)

model <- glm(pa ~ poultryramsar, family=binomial, data=dIN, x=T,y=T)
model2<- lrm(pa ~ poultryramsar,  data=dIN, x=T,y=T)
### Extracting model information
output[i,"AIC"]=AIC(model)
output[i,"R2"]<- model2$stats[10]
output[i,"P-Value"]<- as.data.frame(anova(model2))$P[2]
a<-length(rownames(summary(model)$coefficients))

for (r in 2:a) {
vname<-rownames(summary(model)$coefficients)[r]
vname2=paste("P_",vname,sep="")
output[i,vname2]<-summary(model)$coefficients[r,4]
vname2=paste("EXP-B_",vname,sep="")
output[i,vname2]<-exp(summary(model)$coefficients[r,1])
vname2=paste("CIExpB-Low_",vname,sep="")
output[i,vname2]<-exp(summary(model)$coefficients[r,1]-1.96*summary(model)$coefficients[r,2])
vname2=paste("CIExpB-Up_",vname,sep="")
output[i,vname2]<-exp(summary(model)$coefficients[r,1]+1.96*summary(model)$coefficients[r,2])
}
#AUC
pr<-predict.glm(model,newdata=dIN, type='response')
dfauc<-as.data.frame(cbind(plotID=1:length(dIN$pa),Observed=dIN$pa,Predicted=pr))
output[i,"AUC"]<-auc(dfauc)[1]
output[i,"AUC_UCI"]<-auc(dfauc)[1]+1.96*auc(dfauc)[2]
output[i,"AUC_LCI"]<-auc(dfauc)[1]-1.96*auc(dfauc)[2]

## Max KAPPA
output[i,"Threshold"] <- optimal.thresholds(dfauc)$Predicted[4]
output[i,"Kappa"]<-Kappa(cmx(dfauc, threshold = output[i,"Threshold"]))[1]
output[i,"Kappa_sd"]<-Kappa(cmx(dfauc, threshold = output[i,"Threshold"]))[2]
}

summary(output)

write.csv(output,"poultryramsar.csv")

###############################################################################
## Univariate analysis ~ square effect
## change 19 times when use another variable

setwd("D:\\5Poultry\\logic\\1univariate") 

bootnum=1000

output<-data.frame(matrix(nrow=bootnum,ncol=17))

names(output)=c('EXP-B_NDVI12','CIExpB-Low_NDVI12','CIExpB-Up_NDVI12','P_NDVI12','EXP-B_SNDVI12','CIExpB-Low_SNDVI12','CIExpB-Up_SNDVI12','P_SNDVI12',
'R2','AIC',"AUC","AUC_LCI","AUC_UCI",'P-Value','Threshold','Kappa','Kappa_sd')

d <- read.csv("environment_com.csv")

for (i in 1:bootnum){

P<-subset(d,select=c(names(d)),d$pa==1)
A<-subset(d,select=c(names(d)),d$pa==0)

Arind<-sample(1:nrow(A),296,replace=T)
Prind<-sample(1:nrow(P),296,replace=T)
Ain<-A[Arind,]
Pin<-P[Prind,]
dIN<- rbind(Ain,Pin)

NDVI12 <- (dIN$NDVI12- mean(d$NDVI12))/sd(d$NDVI12)
SNDVI12<- NDVI12^2

model <- glm(pa ~ NDVI12 + SNDVI12, family=binomial, data=dIN, x=T,y=T)
model2<- lrm(pa ~ NDVI12 + SNDVI12,  data=dIN, x=T,y=T)

### Extracting model information
output[i,"AIC"]=AIC(model)
output[i,"R2"]<- model2$stats[10]
output[i,"P-Value"]<- as.data.frame(anova(model2))$P[2]
a<-length(rownames(summary(model)$coefficients))

for (r in 2:a) {
vname<-rownames(summary(model)$coefficients)[r]
vname2=paste("P_",vname,sep="")
output[i,vname2]<-summary(model)$coefficients[r,4]
vname2=paste("EXP-B_",vname,sep="")
output[i,vname2]<-exp(summary(model)$coefficients[r,1])
vname2=paste("CIExpB-Low_",vname,sep="")
output[i,vname2]<-exp(summary(model)$coefficients[r,1]-1.96*summary(model)$coefficients[r,2])
vname2=paste("CIExpB-Up_",vname,sep="")
output[i,vname2]<-exp(summary(model)$coefficients[r,1]+1.96*summary(model)$coefficients[r,2])
}
#AUC
pr<-predict.glm(model,newdata=dIN, type='response')
dfauc<-as.data.frame(cbind(plotID=1:length(dIN$pa),Observed=dIN$pa,Predicted=pr))
output[i,"AUC"]<-auc(dfauc)[1]
output[i,"AUC_UCI"]<-auc(dfauc)[1]+1.96*auc(dfauc)[2]
output[i,"AUC_LCI"]<-auc(dfauc)[1]-1.96*auc(dfauc)[2]

## Max KAPPA
output[i,"Threshold"] <- optimal.thresholds(dfauc)$Predicted[4]
output[i,"Kappa"]<-Kappa(cmx(dfauc, threshold = output[i,"Threshold"]))[1]
output[i,"Kappa_sd"]<-Kappa(cmx(dfauc, threshold = output[i,"Threshold"]))[2]
}

summary(output)

write.csv(output,"NDVI122.csv")

###############################################################################
## Moran's I ~ only significant varaibles
## [12] "dem" [13] "slope"   [15] "popden" [17]"mapet"  
## [19]"tmin01" [20]"tmin02" [21]"tmin03"  [22]"tmin04" [23]"tmin05" [24]"tmin06"    
## [25]"tmin07" [26]"tmin08" [27]"tmin09" [28]"tmin10" [29]"tmin11" [30]"tmin12"    
## [31]"tmax01" [32]"tmax02"  [33]"tmax03" [34]"tmax04" [35]"tmax05" [36]"tmax06"    
## [37]"tmax07" [38]"tmax08" [39]"tmax09" [40]"tmax10" [41]"tmax11" [42]"tmax12"    
## [43]"pre01" [44]"pre02" [45]"pre03" [46]"pre04" [47]"pre05" [48]"pre06" 
## [49]"pre07" [50]"pre08" [51]"pre09" [52]"pre10" [53]"pre11" [54]"pre12" 
## [55]"NDVI01" [58]"NDVI04" [59]"NDVI05" [60]"NDVI06" [61]"NDVI07" 
## [62]"NDVI08" [63]"NDVI09" [64]"NDVI10" [65]"NDVI11"
## [67]"popdenglwd"  [68]"popdenramsar" [69] "poultryglwd" 

setwd("D:\\5Poultry\\logic\\2Moran")
##IMPORT THE DATASET
library(foreign)
library(ncf)
library(spdep)

d <- read.csv("environment_com.csv")

# change variable here
var=55
names(d[var])
bootnum=1000
range=10

##SUBSETTING into 2 datasets (presence (P) and absence (A))
A<-subset(d,select=c(names(d)),d$pa==0)
P<-subset(d,select=c(names(d)),d$pa==1)

outputmoran <-data.frame(matrix(nrow=bootnum,ncol=4))
names(outputmoran)=c("MoranI","Expectation","Variance","P-Value")

##BOOTSTRAP          
for(n in 1:bootnum)
{
Arind<-sample(1:nrow(A),130,replace=T)
Prind<-sample(1:nrow(P),130,replace=T)
Ain<-A[Arind,]
Pin<-P[Prind,]
dIN<- rbind(Ain,Pin)
nbS<-dnearneigh(as.matrix(dIN[2:3]),0,range)
nbS.listw<-nb2listw(nbS)

mor<-moran.test(dIN[,var],listw=nbS.listw)

outputmoran[n,1]<-mor[3]$estimate[1]
outputmoran[n,2]<-mor[3]$estimate[2]
outputmoran[n,3]<-mor[3]$estimate[3]
outputmoran[n,4]<-mor[2]$p.value
  #calculate moran's I
}
summary (outputmoran)

write.csv(outputmoran,file="NDVI01.csv",row.names=F)

###############################################################################
## VIF

library(rms)  
library(PresenceAbsence)

setwd("D:\\5Poultry\\logic\\3VIF")
 
d <- read.csv("environment_com.csv")

bootnum=1000

output<-data.frame(matrix(nrow=bootnum,ncol=9)
names(output)=c("dem","tmin04" ,"tmax11", "pre04", "pre09","pre11","NDVI01","NDVI04","NDVI09","popdenglwd")

for (i in 1:bootnum){                        

P<-subset(d,select=c(names(d)),d$pa==1)
A<-subset(d,select=c(names(d)),d$pa==0)

Arind<-sample(1:nrow(A),130,replace=T)
Prind<-sample(1:nrow(P),130,replace=T)
Ain<-A[Arind,]
Pin<-P[Prind,]
dIN<- rbind(Ain,Pin)

model <- glm(pa ~ dem+tmin04+tmax11
+pre04+pre09+pre11+NDVI01+NDVI04+NDVI09+popdenglwd, 
data=dIN, x=T,y=T)
a<-vif(model)

output[i,'dem']<- a[1]
output[i,'tmin04']<- a[2]
output[i,'tmax11']<- a[3]
output[i,'pre04']<- a[4]
output[i,'pre09']<- a[5]
output[i,'pre11']<- a[6]
output[i,'NDVI01']<- a[7]
output[i,'NDVI04']<- a[8]
output[i,'NDVI09']<- a[9]
output[i,'popdenglwd']<- a[10]

 }
summary(output)

write.csv(output,"vif10.csv")

################################################################################
##stepwise selection
##13 variables:'dem','pre04','pre09','pre11','Spre11', 'tmin04','tmax11', 'Stmax11'
##'NDIVI01','SNDVI01','NDVI04', 'NDVI09', 'popdenglwd'

library(MASS)
library(rms)
library(PresenceAbsence)
##############################

setwd("D:\\5poultry\\logic\\4multivariate")

bootnum=1000
matrixAUC.x<-matrix(nrow=101,ncol=bootnum)
matrixAUC.y<-matrix(nrow=101,ncol=bootnum)

output<-data.frame(matrix(nrow=bootnum,ncol=77))
names(output)=c('AIC','NO_Selected',
'C_Intercept','C_dem', 'C_pre04','C_pre09','C_pre11','C_Spre11','C_tmin04','C_tmax11',
'C_Stmax11','C_NDVI01','C_SNDVI01','C_NDVI04','C_NDVI09','C_popdenglwd',
'P_Intercept','P_dem','P_pre04','P_pre09','P_pre11','P_Spre11', 'P_tmin04','P_tmax11',
'P_Stmax11','P_NDVI01','P_SNDVI01','P_NDVI04','P_NDVI09','P_popdenglwd',
'EXP-B_Intercept','EXP-B_dem','EXP-B_pre04','EXP-B_pre09','EXP-B_pre11','EXP-B_Spre11', 
'EXP-B_tmin04','EXP-B_tmax11','EXP-B_Stmax11','EXP-B_NDVI01','EXP-B_SNDVI01',
'EXP-B_NDVI04','EXP-B_NDVI09','EXP-B_popdenglwd',
'CIExpB-Low_Intercept','CIExpB-Low_dem','CIExpB-Low_pre04','CIExpB-Low_pre09',
'CIExpB-Low_pre11','CIExpB-Low_Spre11', 'CIExpB-Low_tmin04','CIExpB-Low_tmax11','CIExpB-Low_Stmax11',
'CIExpB-Low_NDVI01','CIExpB-Low_SNDVI01','CIExpB-Low_NDVI04','CIExpB-Low_NDVI09',
'CIExpB-Low_popdenglwd','CIExpB-Up_Intercept','CIExpB-Up_dem','CIExpB-Up_pre04',
'CIExpB-Up_pre09', 'CIExpB-Up_pre11', 'CIExpB-Up_Spre11', 'CIExpB-Up_tmin04','CIExpB-Up_tmax11',
'CIExpB-Up_Stmax11','CIExpB-Up_NDVI01','CIExpB-Up_SNDVI01',
'CIExpB-Up_NDVI04','CIExpB-Up_NDVI09','CIExpB-Up_popdenglwd',
"AUC","AUC_sd", "Threshold", "Kappa", "Kappa_sd")


d <- read.csv("environment_com.csv")

# Z score transformation
dem <- (dIN$dem- mean(d$dem))/sd(d$dem)
pre04 <- (dIN$pre04- mean(d$pre04))/sd(d$pre04)
pre09 <- (dIN$pre09- mean(d$pre09))/sd(d$pre09)
pre11 <- (dIN$pre11- mean(d$pre11))/sd(d$pre11)
Spre11 <- pre11^2
tmin04 <- (dIN$tmin04- mean(d$tmin04))/sd(d$tmin04)
tmax11 <- (dIN$tmax11- mean(d$tmax11))/sd(d$tmax11)
Stmax11 <- tmax11^2
NDVI01 <- (dIN$NDVI01- mean(d$NDVI01))/sd(d$NDVI01)
SNDVI01 <- NDVI01^2 
NDVI04 <- (dIN$NDVI04- mean(d$NDVI04))/sd(d$NDVI04)
NDVI09 <- (dIN$NDVI09- mean(d$NDVI09))/sd(d$NDVI09)
popdenglwd <- (dIN$popdenglwd-mean(d$popdenglwd))/sd(d$popdenglwd)

for (i in 1:bootnum){

P<-subset(d,select=c(names(d)),d$pa==1)
A<-subset(d,select=c(names(d)),d$pa==0)

Arind<-sample(1:nrow(A),130,replace=T)
Prind<-sample(1:nrow(P),130,replace=T)
Ain<-A[Arind,]
Pin<-P[Prind,]
dIN<- rbind(Ain,Pin)

model <- glm(pa ~ dem+pre04+pre09+pre11+Spre11+tmin04+tmax11+Stmax11+NDVI01+SNDVI01
+NDVI04+NDVI09+popdenglwd, family= binomial, data=dIN, x=T,y=T)
modstep<-stepAIC(model)

### Extracting model information
output[i,'AIC']<-AIC(modstep)

a<-length(rownames(summary(modstep)$coefficients))
output[i,'NO_Selected']<-a-1

output[i,'C_Intercept']<-summary(modstep)$coefficients[1,1]
output[i,'P_Intercept']<-summary(modstep)$coefficients[1,4]
output[i,'EXP-B_Intercept']<-exp(summary(modstep)$coefficients[1,1])
output[i,'CIExpB-Low_Intercept']<-exp(summary(modstep)$coefficients[1,1]-1.96*summary(modstep)$coefficients[1,2])
output[i,'CIExpB-Up_Intercept']<-exp(summary(modstep)$coefficients[1,1]+1.96*summary(modstep)$coefficients[1,2])

for (r in 2:a) {
vname<-rownames(summary(modstep)$coefficients)[r]
vname2=paste("C_",vname,sep="")
output[i,vname2]<-summary(modstep)$coefficients[r,1]
vname2=paste("P_",vname,sep="")
output[i,vname2]<-summary(modstep)$coefficients[r,4]
vname2=paste("EXP-B_",vname,sep="")
output[i,vname2]<-exp(summary(modstep)$coefficients[r,1])
vname2=paste("CIExpB-Low_",vname,sep="")
output[i,vname2]<-exp(summary(modstep)$coefficients[r,1]-1.96*summary(modstep)$coefficients[r,2])
vname2=paste("CIExpB-Up_",vname,sep="")
output[i,vname2]<-exp(summary(modstep)$coefficients[r,1]+1.96*summary(modstep)$coefficients[r,2])
}


pr<-predict.glm(modstep,newdata=dIN, type='response')
dfauc<-as.data.frame(cbind(plotID=1:length(dIN$pa),Observed=dIN$pa,Predicted1=pr))
output[i,"AUC"]<-auc(dfauc)[1]
output[i,"AUC_SD"]<-auc(dfauc)[2]

#write.csv(dfauc,paste("dfauc",as.character(i),".csv",sep=""))
b<-roc.plot.calculate(dfauc)
matrixAUC.x[,i]<-1-b[,4]
matrixAUC.y[,i]<-b[,3]
}

write.csv(output,"Stepwise_selection.csv")

################################################################################
## multiple logistic regression ~ final model 
##4 variables: 'pre11', 'tmax11','Stmax11', 'popdenglwd'

library(MASS)
library(rms)
library(PresenceAbsence)


setwd("D:\\5poultry\\logic\\4multivariate")

bootnum=1000
matrixAUC.x<-matrix(nrow=101,ncol=bootnum)
matrixAUC.y<-matrix(nrow=101,ncol=bootnum)

##'pre11', 'tmax11','Stmax11', 'popdenglwd'

output<-data.frame(matrix(nrow=bootnum,ncol=32))
names(output)=c('AIC','NO_Selected',
'C_Intercept','C_pre11','C_tmax11','C_Stmax11','C_popdenglwd',
'P_Intercept','P_pre11', 'P_tmax11','P_Stmax11','P_popdenglwd',
'EXP-B_Intercept','EXP-B_pre11','EXP-B_tmax11','EXP-B_Stmax11','EXP-B_popdenglwd',
'CIExpB-Low_Intercept','CIExpB-Low_pre11','CIExpB-Low_tmax11','CIExpB-Low_Stmax11',
'CIExpB-Low_popdenglwd',
'CIExpB-Up_Intercept','CIExpB-Up_pre11','CIExpB-Up_tmax11',
'CIExpB-Up_Stmax11','CIExpB-Up_popdenglwd',
"AUC","AUC_sd", "Threshold", "Kappa", "Kappa_sd")


d <- read.csv("environment_com.csv")

# Z score transformation

pre11 <- (dIN$pre11- mean(d$pre11))/sd(d$pre11)
tmax11 <- (dIN$tmax11- mean(d$tmax11))/sd(d$tmax11)
Stmax11 <- tmax11^2
popdenglwd <- (dIN$popdenglwd-mean(d$popdenglwd))/sd(d$popdenglwd)

for (i in 1:bootnum){

P<-subset(d,select=c(names(d)),d$pa==1)
A<-subset(d,select=c(names(d)),d$pa==0)

Arind<-sample(1:nrow(A),130,replace=T)
Prind<-sample(1:nrow(P),130,replace=T)
Ain<-A[Arind,]
Pin<-P[Prind,]
dIN<- rbind(Ain,Pin)

model <- glm(pa ~pre11+tmax11+Stmax11+popdenglwd, family= binomial, data=dIN, x=T,y=T)

### Extracting model information
output[i,'AIC']<-AIC(model)

a<-length(rownames(summary(model)$coefficients))
output[i,'NO_Selected']<-a-1

output[i,'C_Intercept']<-summary(model)$coefficients[1,1]
output[i,'P_Intercept']<-summary(model)$coefficients[1,4]
output[i,'EXP-B_Intercept']<-exp(summary(model)$coefficients[1,1])
output[i,'CIExpB-Low_Intercept']<-exp(summary(model)$coefficients[1,1]-1.96*summary(model)$coefficients[1,2])
output[i,'CIExpB-Up_Intercept']<-exp(summary(model)$coefficients[1,1]+1.96*summary(model)$coefficients[1,2])

for (r in 2:a) {
vname<-rownames(summary(model)$coefficients)[r]
vname2=paste("C_",vname,sep="")
output[i,vname2]<-summary(model)$coefficients[r,1]
vname2=paste("P_",vname,sep="")
output[i,vname2]<-summary(model)$coefficients[r,4]
vname2=paste("EXP-B_",vname,sep="")
output[i,vname2]<-exp(summary(model)$coefficients[r,1])
vname2=paste("CIExpB-Low_",vname,sep="")
output[i,vname2]<-exp(summary(model)$coefficients[r,1]-1.96*summary(model)$coefficients[r,2])
vname2=paste("CIExpB-Up_",vname,sep="")
output[i,vname2]<-exp(summary(model)$coefficients[r,1]+1.96*summary(model)$coefficients[r,2])
}

pr<-predict.glm(model,newdata=dIN, type='response')
dfauc<-as.data.frame(cbind(plotID=1:length(dIN$pa),Observed=dIN$pa,Predicted1=pr))
output[i,"AUC"]<-auc(dfauc)[1]
output[i,"AUC_sd"]<-auc(dfauc)[2]

## Max KAPPA
output[i,"Threshold"] <- optimal.thresholds(dfauc)$Predicted[4]
output[i,"Kappa"]<-Kappa(cmx(dfauc, threshold = output[i,"Threshold"]))[1]
output[i,"Kappa_sd"]<-Kappa(cmx(dfauc, threshold = output[i,"Threshold"]))[2]

#write.csv(dfauc,paste("dfauc",as.character(i),".csv",sep=""))
b<-roc.plot.calculate(dfauc)
matrixAUC.x[,i]<-1-b[,4]
matrixAUC.y[,i]<-b[,3]

}

summary(output)

write.csv(output,"finalmodel_04.csv")

write.csv(matrixAUC.x,"MatAUCX_04.csv")
write.csv(matrixAUC.y,"MatAUCY_04.csv")