Untitled

#TRATAMENTO DE DADOS
dados <- read.delim("Trabalho10_Dados.txt", header= FALSE, sep="")

#escolha dos dados pedidos no enunciado
dados <- dados[,2:13]
names(dados)<- c("Age", "Height", "Sex", "Survival", "Shock.Type", "Systolic.Pressure", 
                 "Mean.Arterial.Pressure", "Heart.Rate", "Diastolic.Pressure", "Mean.Central.Venous.Pressure", 
                 "Body.Surface.Area", "Cardiac.Index")
rows <- nrow(dados)
even_rows <- seq_len(rows) %% 2
dados_inicial <- dados[even_rows == 1,]
rownames(dados_inicial) <- NULL #renumerar linhas
nrows <- nrow(dados_inicial)

#transformar qualitativas
qualitativas <- c("Sex", "Survival", "Shock.Type")
for (i in 1:ncol(dados_inicial)){
  if (names(dados_inicial[i]) %in% qualitativas){
    dados_inicial[,i] <- factor(dados_inicial[,i])
  }
}
#ANALISE DE DADOS
summary(dados_inicial)

#box.plot continuas
dados_cont <- dados_inicial[ ,-c(match(qualitativas, names(dados_inicial)))]
par(mfrow = c(3, ncol(dados_cont)/3), mar=c(2,2,2,2), cex=0.5)
lapply(1:ncol(dados_cont), function(i) boxplot(dados_cont[,i], main=names(dados_cont)[i])) 

#variavel resposta
y <- matrix(dados_inicial$Cardiac.Index)
dados_inicial <- dados_inicial[,-ncol(dados_inicial)]

#correlações
library(corrplot)
a <- cor(data.matrix(dados_cont))
dev.off()
corrplot(a, method = 'color', addCoef.col = 'black', 
         number.cex = 0.45, tl.cex = 0.5, tl.col="black")

#eliminar Mean Arterial
mrl.comp <-lm(y ~ .,data=dados_inicial)
library(car)
library(usdm)
vif(mrl.comp)
dados_inicial <- dados_inicial[, -which(colnames(dados_inicial) %in% c("Mean.Arterial.Pressure"))]
mrl.comp <-lm(y ~ .,data=dados_inicial)
vif(mrl.comp)


#ANALISE PRELIMINAR DO MODELO COMPLETO
#modelo completo com covariaveis eliminadas
#plot(mrl.comp)
summary(mrl.comp)
extractAIC(mrl.comp)

library(gvlma)
gvlma(mrl.comp)


#boxcox para por y normal
library(MASS)
b <- boxcox(lm(y ~ 1))
lambda <- b$x[which.max(b$y)] #lambda=0.30303...
y2 <- (y^lambda-1)/lambda

#teste de ajustamento do y
shapiro.test(y) #p-value = 1.101e-05
shapiro.test(y2) #p-value = 0.7035 > 0.25


#modelo com y normal
mrl.comp2 <-lm(y2 ~ .,data=dados_inicial)
#plot(mrl.comp2) #mais proximo de normal

#eliminar outliers - nao tem pelo box plot 
boxplot(y2)

##dados de treino e de teste
set.seed(73)             
teste_ind<-sort(sample(nrows,0.2*nrows))

#Treino
dados_treino <-dados_inicial[-teste_ind,]
y_treino <- y2[-teste_ind]
summary(dados_treino)
shapiro.test(y_treino)

#Teste
dados_teste <- dados_inicial[teste_ind,]
y_teste <- y2[teste_ind]
shapiro.test(y_teste)

#matrizes x 
X <- cbind(1,data.matrix(dados_treino))
n=length(y_treino)

mrl.full <- lm(y_treino~.,data=dados_treino)
summary(mrl.full)
extractAIC(mrl.full)

#stepforward
#primeira int
mrl.base <-lm(y_treino~1,
            data=dados_treino)

mrl.stepforward <- step(mrl.base,
                    scope = list(upper = formula(mrl.full), 
                                 lower = formula(mrl.base)),
                    direction = "forward", trace="FALSE")

summary(mrl.stepforward)
extractAIC(mrl.stepforward)
formula(mrl.stepforward)

#segunda int
mrl.int2 <- lm(y_treino ~ (Shock.Type + Sex + Diastolic.Pressure + Systolic.Pressure + 
                             Age + Body.Surface.Area)^2, 
               data = dados_treino)

mrl.stepforward2 <- step(mrl.stepforward,
                         scope = list(upper = formula(mrl.int2), 
                                      lower = formula(mrl.stepforward)),
                         direction = "forward", trace="FALSE")
summary(mrl.stepforward2)
extractAIC(mrl.stepforward2)
formula(mrl.stepforward2)

#terceira int
mrl.int3 <- update(mrl.stepforward2, .~. + Sex:Body.Surface.Area:Diastolic.Pressure)

mrl.stepforward3 <- step(mrl.stepforward2,
                         scope = list(upper = formula(mrl.int3), 
                                      lower = formula(mrl.stepforward2)),
                         direction = "forward", trace="FALSE")
summary(mrl.stepforward3)
extractAIC(mrl.stepforward3)
formula(mrl.stepforward3)


library(ggplot2)
predictions2 <- predict(mrl.stepforward, newdata = dados_teste)
d2<-data.frame(predictions2, y=y_teste)
ggplot(d2, aes(predictions2, y)) +
  geom_point(shape = 16, size = 3, show.legend = FALSE) 

ri <- y_teste - predictions2
sum(ri^2)/length(y_teste)

#bestsubset com r^2 ajustado
library(leaps)
bestsubsets <- regsubsets(y_treino~., data = dados_treino, nvmax = ncol(dados_treino))  
plot(bestsubsets, scale = "adjr2")
bs_summary <- summary(bestsubsets) 
ind_r2 <- which.max(bs_summary$adjr2)
best.r<- bestsubsets[ind_r2]
coef.r <- best.r$thetab
coef.r <- coef(bestsubsets, ind_r2)

mrl.best.r<- lm(y_treino ~ ., data = dados_treino)

summary(mrl.best.r)
extractAIC(mrl.best.r)
formula(mrl.best.r)



plot(bs_summary$adjr2, xlab = "Number of Variables", ylab = "Adjusted RSq", type = "b")
best_adj_r2 = which.max(bs_summary$adjr2)
points(best_adj_r2, bs_summary$adjr2[best_adj_r2],
       col = "red", cex = 2, pch = 20)



#bestsubset com aic
bestsubsets2 <- best.subset(y_treino~., data = dados_treino, method="aic")





#################

#stepbackward
mrl.stepbackward <- step(mrl.full, direction = "backward", trace=FALSE)

summary(mrl.stepbackward)
extractAIC(mrl.stepbackward)
formula(mrl.stepbackward)
  
  
library(ggplot2)
predictions <- predict(mrl.full, newdata = dados_treino)
d<-data.frame(predictions, y=y_treino)
ggplot(d, aes(predictions, y)) +
  geom_point(shape = 16, size = 3, show.legend = FALSE) 







library(emmeans)

emmip(mrl.stepbackward,Sex ~ Age, cov.reduce=range) #não há interação
emmip(mrl.stepbackward,Shock.Type ~ Age, cov.reduce=range) #não há interação
emmip(mrl.full,Sex ~ Body.Surface.Area, cov.reduce=range) #não há interação
emmip(mrl.full,Age ~ Height, cov.reduce=range) #não há interação