setwd("C:\\Users\\srijana\\Desktop\\RcurrentProject")

Nut<- read.csv("Nutrients_plusOne.csv")
attach(Nut)
detach(Nut)

Nut$WH.ratio <-Waist_Cir/Hip_Cir
Nut$WH.ratioCat <- ifelse(Sex==2&Nut$WH.ratio<0.95|Sex==1&Nut$WH.ratio<0.8,1,2)
Nut$BMIcat <- ifelse(Nut$BMI<27.5,1,2)
CAD_group<-as.factor(CAD_group)
Smoking <- as.factor(Smoking)
Alcohol <- as.factor(Alcohol)
Dyslipidemia <- as.factor(Dyslipidemia)
T2DM <- as.factor(T2DM)
Nut$WH.ratioCat <- as.factor(Nut$WH.ratioCat)
Physical.activity <- as.factor(Physical.activity)


# calculation for sociodemographic variable (Table 1)
summary(Age)
summary(Age[Nut$CAD_group==0])
summary(Age[Nut$CAD_group==1])
wilcox.test(age1,age2,paired = TRUE,correct = FALSE)
wilcox.test(Age[Nut$CAD_group==0],
            Age[Nut$CAD_group==1],paired = TRUE,correct = FALSE)

mcnemar.test(T2DM[Nut$CAD_group==0],T2DM[Nut$CAD_group==1],correct = FALSE)
table(T2DM,CAD_group)

mcnemar.test(Hypertension[CAD_group==0],Hypertension[CAD_group==1],correct = FALSE)
table(Hypertension,CAD_group)

table(Dyslipidemia, CAD_group)
mcnemar.test(Dyslipidemia[Nut$CAD_group==0],
             Dyslipidemia[Nut$CAD_group==1],correct = FALSE)

table(Nut$BMIcat,CAD_group)
mcnemar.test(Nut$BMIcat[Nut$CAD_group==0],
             Nut$BMIcat[Nut$CAD_group==1],correct = FALSE)

table(Nut$WH.ratioCat,CAD_group)
mcnemar.test(Nut$WH.ratioCat[Nut$CAD_group==0],
             Nut$WH.ratioCat[Nut$CAD_group==1],correct = FALSE)

table(Alcohol, CAD_group)
mcnemar.test(Alcohol[Nut$CAD_group==0],
             Alcohol[Nut$CAD_group==1],correct = FALSE)

table(Smoking, CAD_group)
mcnemar.test(Smoking [Nut$CAD_group==0],
            Smoking [Nut$CAD_group==1],correct = FALSE)

table(Physical.activity, CAD_group)
mcnemar.test(Physical.activity[Nut$CAD_group==0],
             Physical.activity[Nut$CAD_group==1],correct = FALSE)

#Nutrients variable (Table 2, S2 Table)
tapply(Food.energy,CAD_group, sd)
Fke1 <- Food.energy [CAD_group==1]
summary(Fke1)
Fke2 <- Food.energy[CAD_group==0]
summary(Fke2)
wilcox.test(Fke1,Fke2, paired = TRUE,correct = FALSE)
t.test(Fke1,Fke2,paired = TRUE)

tapply(Protein,CAD_group,sd)
prt1 <-Protein[CAD_group==0]
summary(prt1)
prt2 <-Protein[CAD_group==1]
summary(prt2)
wilcox.test(prt1,prt2,paired = TRUE,correct = FALSE)
t.test(prt1,prt2,paired = TRUE)

tapply(Total.fat.oil,CAD_group,sd)
lpd1<- Total.fat.oil[CAD_group==0]
summary(lpd1)
lpd2 <- Total.fat.oil[CAD_group==1]
summary(lpd2)
wilcox.test(lpd1,lpd2,paired = TRUE,correct=FALSE)
t.test(lpd1,lpd2,paired = TRUE)

tapply(Carbohydrate,CAD_group,sd)
carb1 <-Carbohydrate[CAD_group==0]
summary(carb1)
carb2 <- Carbohydrate[CAD_group==1]
wilcox.test(carb1,carb2,paired = TRUE)
summary(carb2)
t.test(carb1,carb2,paired = TRUE)

tapply(Fiber,CAD_group,sd)
fbr1 <- Fiber[CAD_group==0]
summary(fbr1)
fbr2 <- Fiber[CAD_group==1]
summary(fbr2)
wilcox.test(fbr1,fbr2,paired = TRUE,correct = FALSE)
t.test(fbr1,fbr2,paired = TRUE)

tapply(Calcium,CAD_group,sd)
cal1<- Calcium[CAD_group==0]
cal2 <-Calcium[CAD_group==1]
wilcox.test(cal1,cal2,paired = TRUE,correct = FALSE)
summary(cal1)
summary(cal2)
t.test(cal1,cal2,paired = TRUE)

tapply(Phosphorus,CAD_group,sd)
ph1 <-Phosphorus[CAD_group==0]
ph2 <- Phosphorus[CAD_group==1]
wilcox.test(ph1,ph2,paired = TRUE,correct = FALSE)
summary(ph1)
summary(ph2)
t.test(ph1,ph2,paired = TRUE)

tapply(Iron,CAD_group,sd)
in1 <- Iron[CAD_group==0]
in2 <- Iron[CAD_group==1]
wilcox.test(in1,in2,paired = TRUE,correct = FALSE)
summary(in1)
summary(in2)
t.test(in1,in2,paired = TRUE)

tapply(Zinc,CAD_group,sd)
zn1 <- Zinc[CAD_group==0]
zn2 <- Zinc[CAD_group==1]
wilcox.test(zn1,zn2,paired = TRUE,correct = FALSE)
summary(zn1)
summary(zn2)
t.test(zn1,zn2,paired = TRUE)

tapply(Thiamine,CAD_group,sd)
th1 <- Thiamine[CAD_group==0]
th2 <- Thiamine[CAD_group==1]
wilcox.test(th1,th2,paired = TRUE,correct = FALSE)
summary(th1)
summary(th2)
t.test(th1,th2,paired = TRUE)

tapply(Riboflavin,CAD_group,sd)
rb1 <- Riboflavin[CAD_group==0]
rb2 <- Riboflavin[CAD_group==1]
wilcox.test(rb1,rb2,paired = TRUE,correct = FALSE)
summary(rb1)
summary(rb2)
t.test(rb1,rb2,paired = TRUE)

tapply(Niacin,CAD_group,sd)
nc1 <- Niacin[CAD_group==0]
nc2 <- Niacin[CAD_group==1]
wilcox.test(nc1,nc2,paired = TRUE,correct = FALSE)
summary(nc1)
summary(nc2)
t.test(nc1,nc2,paired = TRUE)

tapply(Vitamin.C,CAD_group,sd)
vc1 <- Vitamin.C[CAD_group==0]
vc2 <- Vitamin.C[CAD_group==1]
summary(vc1)
summary(vc2)
wilcox.test(vc1,vc2,paired = TRUE,correct = FALSE)
t.test(vc1,vc2,paired = TRUE)

tapply(Beta.carotene,CAD_group,sd)
ct1 <- Beta.carotene[CAD_group==0]
ct2 <- Beta.carotene[CAD_group==1]
wilcox.test(ct1,ct2,paired = TRUE,correct = FALSE)
summary(ct1)
summary(ct2)
t.test(ct1,ct2,paired = TRUE)

tapply(Vitamin.A.R.E.,CAD_group,sd)
va1 <- Vitamin.A.R.E.[CAD_group==0]
va2 <- Vitamin.A.R.E.[CAD_group==1]
wilcox.test(va1,va2,paired = TRUE,correct = FALSE)
summary(va1)
summary(va2)
t.test(va1,va2,paired = TRUE)

tapply(PUFA,CAD_group,sd)
pufa1<- PUFA[CAD_group==0]
pufa2 <- PUFA[CAD_group==1]
summary(pufa1)
summary(pufa2)
wilcox.test(pufa1,pufa2,paired = TRUE,correct = FALSE)
t.test(pufa1,pufa2,paired = TRUE)

tapply(SFA,CAD_group,sd)
sfa1 <- SFA[CAD_group==0]
sfa2 <-SFA[CAD_group==1]
wilcox.test(sfa1,sfa2,paired = TRUE,correct = FALSE)
summary(sfa1)
summary(sfa2)
t.test(sfa1,sfa2,paired = TRUE)

tapply(MUFA,CAD_group,sd)
mufa1 <- MUFA[CAD_group==0]
mufa2 <- MUFA[CAD_group==1]
summary(mufa1)
summary(mufa2)
wilcox.test(mufa1,mufa2,paired = TRUE,correct = FALSE)
t.test(mufa1,mufa2,paired = TRUE)

tapply(Cholesterol,CAD_group,sd)
cl1 <- Cholesterol[CAD_group==0]
cl2 <- Cholesterol[CAD_group==1]
wilcox.test(cl1,cl2,paired = TRUE,correct = FALSE)
summary(cl1)
summary(cl2)
t.test(cl1,cl2,paired = TRUE)

#correlation matrix of nutrients
NedCor <- subset(Nut, select=c(Food.energy,Carbohydrate,Protein,Total.fat.oil,PUFA,MUFA, SFA,Cholesterol,
                               Fiber,Thiamine,Niacin,Riboflavin,
                               Beta.carotene,Vitamin.A.R.E.,Vitamin.C,Zinc,Iron,Calcium,Phosphorus))
round(cor(NedCor),2)
library(rtf)
ColNut <- RTF("rts.doc")
addParagraph(ColNut,"correlation among nutrients\n")
addTable(ColNut,as.data.frame(round(cor(NedCor),2)))
done(ColNut)


library(survival)
library(car)


fitnut1 <- clogit(CAD_group ~ Food.energy+Carbohydrate+Total.fat.oil+Fiber+
                     Thiamine+Riboflavin+Beta.carotene+Vitamin.C+Zinc+Vitamin.A.R.E.
                   +Iron+PUFA+MUFA+SFA+Cholesterol+T2DM+Dyslipidemia+Alcohol+
                     Smoking+Physical.activity+Age+BMI+WH.ratioCat+
                     strata(Pair_CADgroup),data=Nut)
summary(fitnut1)
sqrt(vif(fitnut1))>2

fitnut2 <- update(fitnut1,~.- MUFA)
summary(fitnut2)
anova(fitnut1,fitnut2, test="log-lik")

fitnut3 <- update(fitnut2,~.- Fiber)
summary(fitnut3)
anova(fitnut2,fitnut3, test="log-lik")

fitnut4 <- update(fitnut3,~.- Thiamine)
summary(fitnut4)
anova(fitnut3,fitnut4, test="log-lik")

fitnut5 <- update(fitnut4,~.- Alcohol)
summary(fitnut5)
anova(fitnut4,fitnut5, test="log-lik")

fitnut6 <- update(fitnut5,~.- Age)
summary(fitnut6)
anova(fitnut5,fitnut6, test="log-lik")

fitnut7 <- update(fitnut6,~.- PUFA)
summary(fitnut7)
anova(fitnut6,fitnut7, test="log-lik")

fitnut8 <- update(fitnut7,~.- Vitamin.A.R.E.)
summary(fitnut8)
anova(fitnut7,fitnut8, test="log-lik")

fitnut9 <- update(fitnut8,~.- WH.ratioCat)
summary(fitnut9)
anova(fitnut8,fitnut9, test="log-lik")

fitnut10 <- update(fitnut9,~.- Riboflavin)
summary(fitnut10)
anova(fitnut9,fitnut10, test="log-lik")

fitnut11 <- update(fitnut10,~.- Zinc)
summary(fitnut11)
anova(fitnut10,fitnut11, test="log-lik")

fitnut12 <- update(fitnut11,~.- Physical.activity)
summary(fitnut12)
anova(fitnut11,fitnut12, test="log-lik")

fitnut13 <- update(fitnut12,~.- Cholesterol)
summary(fitnut13)
anova(fitnut12,fitnut13, test="log-lik")


#Energy adjusted final model
fitAll <- clogit(CAD_group ~ Food.energy+Carbohydrate+Total.fat.oil
                 +Beta.carotene+Vitamin.C+Iron+SFA+Cholesterol+Smoking+T2DM+BMI
                 +Dyslipidemia+strata(Pair_CADgroup),data=Nut)

summary(fitAll)

# Random Forest analysis

library(randomForest)

Xnut <- subset(Nut, select = c(T2DM,Dyslipidemia,Smoking,Physical.activity,WH.ratioCat, BMI, Alcohol,Age,
                               Food.energy,Carbohydrate,Total.fat.oil,Fiber,Thiamine,Riboflavin,
                               Vitamin.A.R.E., Beta.carotene,Vitamin.C,Zinc,
                               Iron,MUFA,PUFA,SFA,Cholesterol))

Ynut <- CAD_group

RS <- data.frame(Ynut,Xnut)

set.seed(1234)
Model.NutAll <- randomForest(Ynut ~ ., data=RS)
plot(Model.NutAll)
rfNews()
mtry <- tuneRF(Xnut, Ynut, stepFactor = 0.4,
               ntreeTry = 250, improve = 0.01,trace = TRUE,plot = TRUE)
Model.Nutrients <- randomForest(Ynut ~ ., data=RS,ntree=250,mtry=4,
                                importance=TRUE,proximity=TRUE)

attributes(Model.Nutrients)
Model.Nutrients$importance

hist(treesize(Model.Nutrients),main = "no of trees",col = "blue")

varImpPlot(Model.Nutrients,main = ""
           ,bg="red",col = "blue",pch=18,n.var=12,sort = TRUE)

print(Model.Nutrients)

#Evaluation of logistic model (Table 5)
install.packages("ROCR")
library(ROCR)

pred<- predict(fitAll,Nut)
pred<-prediction(pred,Nut$CAD_group)

eval<- performance(pred,"acc")
plot(eval)
max<-which.max(slot(eval,"y.values")[[1]])
acc <-slot(eval,"y.values")[[1]][max]
acc
cut<-slot(eval,"x.values")[[1]][max]
print(c(Accuracy=acc,Cutoff=cut))
cut
tab<-ifelse(pred>0.29,1,0)
table(tab, Nut$CAD_group)
TPR,recall
264/(264+18)
TNR
288/(288+42)
precision,sensitivity
264/(264+42)
specificity
288/(288+18)
F1-score
2*264/((2*264)+18+42)
(2* 0.8627451*0.9361702)/(0.8627451+0.9361702)

#Evaluation of the RF model
accuracy
(257+259)/(257+259+47+49)

TPR,recall
257/(257+49)
TNR
259/(259+47)
precision, sensitivity
257/(257+47)
specificity
259/(259+49)
f1-score
(2*0.8453947*0.8398693)/(0.8453947+0.8398693)

#ROC curve (Figure 3)
install.packages("pROC")
library(pROC)

par(pty="m")
roc(Nut$CAD_group,predict(fitnut12),
    plot=TRUE,legacy.axes=TRUE, percent=TRUE,
    xlab="1-Specificity", ylab="Sensitivity",
    col="blue",lwd=3,print.auc=TRUE)


plot.roc(Nut$CAD_group, Model.Nutrients$votes[,1],percent=TRUE, col="red", lwd=3,
         add=TRUE, print.auc=TRUE,print.auc.y=40)

legend("bottomright",legend=c("Conditional Logistic Rgression", "Random Forest Regression"),
       col=c("blue","red"),lwd = 3,cex=0.82)

#Train Test data
set.seed(1234)
Nut1 <- Nut[CAD_group==1,]
Nut0 <- Nut[CAD_group==0,]
ind <- sample(2,nrow(Nut1),replace=TRUE,prob=c(0.7,0.3))

training <- Nut1[ind==1,]
testing <-Nut1[ind==2,]
testing$Pair_CADgroup


testC<-Nut0[c(5,  14,  16,  26,  28,  29,  41,  37,  39,  50,  56,  58,  60,  61,  66,  72,  74,  81,  86,  90,  92,  99,
              111, 113, 116, 117, 119, 120, 121, 123, 124, 131, 135, 136, 140, 141, 147, 149, 154, 155, 157, 166, 170, 172,
              175, 183, 185, 193, 190, 192, 194, 195, 196, 197, 197, 202, 204, 208, 211, 213, 216, 218, 224, 227, 238, 232,
              235, 241, 243, 247, 248, 259, 269, 267, 270, 273, 276, 280, 283, 284, 290, 294, 296, 302, 305),]

trainC<-Nut0[-c(5,  14,  16,  26,  28,  29,  41,  37,  39,  50,  56,  58,  60,  61,  66,  72,  74,  81,  86,  90,  92,  99,
                111, 113, 116, 117, 119, 120, 121, 123, 124, 131, 135, 136, 140, 141, 147, 149, 154, 155, 157, 166, 170, 172,
                175, 183, 185, 193, 190, 192, 194, 195, 196, 197, 197, 202, 204, 208, 211, 213, 216, 218, 224, 227, 238, 232,
                235, 241, 243, 247, 248, 259, 269, 267, 270, 273, 276, 280, 283, 284, 290, 294, 296, 302, 305),]


train<-rbind(training,trainC)
test<-rbind(testing, testC)
dim(train)

#Note: It does not work because of less sample size