Institut für Medizinische Biometrie
und Statistik (IMBI)

Section 3.2

First simulation example: with constant cause-specifc hazards.

• set.seed(1383) event.times <- rexp(100, 0.9) f.cause <- rbinom(100, size = 1, prob = 1/3) f.cause <- ifelse(f.cause == 0, 2, 1) cens.times <- runif(100,0,5) ## truncation times lt.times <- rgamma(100, shape= 0.5, rate = 2) obs.times <- pmin(event.times, cens.times) obs.cause <- c(event.times <= cens.times) * f.cause table(obs.cause)

• ## Drawing of uniformly distributed random variables my.times <- runif(100) ## inversion method my.times <- -0.5 + (0.25 - log(1 - my.times))^0.5 ## function that computes the cause of failure at time x cause1 <- function(x) { out <- rbinom(length(x), 1, prob = 1 / (1 + 2 * x)) ifelse(out == 0, 2, 1) } my.cause <- cause1(my.times)

• generate.my.times.v2 <- function(n, max.int) { temp <- function(x,y) { return(x + x^2 + y) } stime <- NULL i <- 1 while(length(stime) < n) { u <- runif(1) ## If endpoints are of opposite sign, call uniroot: if (temp(0, log(1 - u)) * temp(max.int, log(1 - u)) < 0) { res <- uniroot(temp, c(0, max.int), tol = 0.0001, y = log(1 - u)) stime[i] <- res$root i <- i + 1 } else cat("Values at endpoints not of opposite signs. \n") } return(stime) }

• my.times <- generate.my.times.v2(n=100, max.int=99)