# PRACTICAL EXERCISE 5
# ====================

# Read bonemarrow transplantation data:
bonemarrow=read.table("http://www.uio.no/studier/emner/matnat/math/STK4080/h12/bone-marrow.txt",header=T)

# We use the mstate package, so this has to be installed and loaded.


# We first define the possible transitions between the three states 
# state 1: transplanted, state 2: platelets recovered, and 3: dead
tmat=transMat(list(c(2, 3), c(3), c()),	names = c("transplanted", "recovered", "relapsed.dead"))


# We then convert the data-frame to long format, where there are 2-3 lines 
# for each person (depending on the trasitions they make):
bonemarrowlong=msprep(time=c(NA,"TP","T2"), status=c(NA,"P","DF"), data=bonemarrow,trans=tmat)

# We then fit a stratified cox-model with no covariates, extract the 
# cumulative transition intensities and make a plot of these
cox.bm=coxph(Surv(Tstart,Tstop,status)~strata(trans),data=bonemarrowlong, method="breslow")
haz.bm=msfit(cox.bm,trans=tmat)
plot(haz.bm,xlab="Days post-transplant",xlim=c(0,1000))

# We the obtain the estimated transition probabilities
prob.bm=probtrans(haz.bm,predt=0)

# We may list the estimated transition probabilities from state 1 by:
prob.bm[[1]]

# And similarily from state 2:
prob.bm[[2]]

# The transition probabilities may be plotted in a stacked manner: 
plot(prob.bm,type="stacked",ord=c(2,1,3))

# Or they may be plotted as separate estimates:
plot(prob.bm,type="single",xlim=c(0,1000))


# To compute estimates with (e.g.) s = 14 you give the command:
prob.bm=probtrans(haz.bm,predt=14)

# The plotting comands remain unchanged