Survival Analysis: Solutions (Part 3)

###############
#             #
# Exercise 1  #
#             #
###############
library(survival)
library(survminer)
data("lung")
lung$status <- as.factor(lung$status) # 1 Censored 2 Dead
final_model <- coxph(Surv(time, status == 2) ~ sex + ph.karno + wt.loss + sex*ph.karno, data = lung)

###############
#             #
# Exercise 2  #
#             #
###############
marting_res <- residuals(final_model, type = "martingale")
residual_df <- data.frame(index = 1:length(marting_res),
                          martingale <- marting_res)

###############
#             #
# Exercise 3  #
#             #
###############
library(ggplot2)
library(dplyr)
residual_df %>%
  ggplot(aes(x = index, y = martingale)) +
  geom_point(alpha = 0.7, colour = "blue") +
  theme_bw()

###############
#             #
# Exercise 4  #
#             #
###############
residual_df$deviance <- residuals(final_model, type = "deviance")
residual_df %>%
  ggplot(aes(x = index, y = deviance)) +
  geom_point(colour = "blue", alpha = 0.7) +
  theme_bw()

###############
#             #
# Exercise 5  #
#             #
###############
library(ggrepel)
residual_df$linPred <- final_model$linear.predictors
residual_df %>%
  ggplot(aes(x = linPred, y = deviance)) +
  geom_point(colour = "blue", alpha = 0.7) +
  geom_text_repel(data = subset(residual_df, linPred > 1),
                  aes(linPred, deviance, label = index)) +
  labs(x = "Linear Predictor Value", y = "Deviance Residual") +
  theme_bw()

###############
#             #
# Exercise 6  #
#             #
###############
dfbetas <- data.frame(residuals(final_model, type = "dfbetas"))

###############
#             #
# Exercise 7  #
#             #
###############
terms <- c("Sex", "ph.Karno", "Weight Loss", "Sex * ph.Karno")
par(mfrow = c(1,2))
for (i in 1:dim(dfbetas)[2]){
  plot(dfbetas[, i], type = "h", col = "blue", ylab = paste("DFBetas for", terms[i]))
}

par(mfrow = c(1,1))

###############
#             #
# Exercise 8  #
#             #
###############
t <- cox.zph(final_model)

###############
#             #
# Exercise 9  #
#             #
###############
plot(t)

################
#              #
# Exercise 10  #
#              #
################
ggsurvplot(survfit(final_model), colour = "blue", ggtheme = theme_bw())