#################################################################################
# Bayesian model selection and statistical modeling 
# Chapman & Hall/CRC Taylor and Francis Group 
#
# Chapter5: Section5.5.4
#
# Survival analysis
#
# Author : Tomohiro Ando
#
################################################################################

library(survival)

data(ovarian)
D <- as.matrix(ovarian)

E <- D[,-(1:2)]; E[,2:4] <- E[,2:4] - 1
y <- as.vector(D[,1])/100
z <- as.vector(D[,2])

n <- length(y)

# Fit exponential model

fit1 <- survreg(Surv(y,z) ~E[,1]+E[,2]+E[,3]+E[,4], dist="exponential")
fit1 <- step(fit1)

BIC1 <- -2*( (fit1$loglik)[2] )+log(n)*length(fit1$coef)


# Fit weibull model

fit2 <- survreg(Surv(y,z) ~E[,1]+E[,2]+E[,3]+E[,4], dist="weibull")
fit2 <- step(fit2)

BIC2 <- -2*( (fit2$loglik)[2] )+log(n)*length(fit2$coef)


# Fit loglogistic model

fit3 <- survreg(Surv(y,z) ~E[,1]+E[,2]+E[,3]+E[,4], dist="loglogistic")
fit3 <- step(fit3)

BIC3 <- -2*( (fit3$loglik)[2] )+log(n)*length(fit3$coef)


# Fit extreme model

fit4 <- survreg(Surv(y,z) ~E[,1]+E[,2]+E[,3]+E[,4], dist="extreme")
fit4 <- step(fit4)

BIC4 <- -2*( (fit4$loglik)[2] )+log(n)*length(fit4$coef)


summary(fit1)
summary(fit2)
summary(fit3)
summary(fit4)

print(BIC1)
print(BIC2)
print(BIC3)
print(BIC4)