Simulate excess lifetime with truncation or right-censoring

This function dispatches simulations accounting for potential left-truncation (remove by setting lower to zero). If type2=ltrt, simulated observations will be lower than the upper bounds upper. If type2=ltrc, simulated observations are capped at upper and the observation is right-censored (rcens=TRUE).

Usage

samp_elife(
  n,
  scale,
  rate,
  shape = NULL,
  lower = 0,
  upper = Inf,
  family = c("exp", "gp", "gomp", "gompmake", "weibull", "extgp", "gppiece",
    "extweibull", "perks", "beard", "perksmake", "beardmake"),
  type2 = c("none", "ltrt", "ltrc", "ditrunc")
)

Arguments

n

sample size

scale

scale parameter(s)

rate

rate parameter(s)

shape

shape parameter(s)

lower

vector of lower bounds

upper

vector of upper bounds

family

string; choice of parametric family

type2

string, either none, ltrt for left- and right-truncated data or ltrc for left-truncated right-censored data

Value

either a vector of observations or, if type2=ltrc, a list with n observations dat and a logical vector of the same length with TRUE for right-censored observations and FALSE otherwise.

Note

As the tails of the Gompertz and Gompertz--Makeham models decrease exponentially fast, the method fails in the rare event case if the lower bound is too large (say larger than the 99.99

Examples

set.seed(1234)
n <- 500L
# Simulate interval truncated data
x <- samp_elife(n = n,
                scale = 2,
                shape = 1.5,
                lower = low <- runif(n),
                upper = upp <- runif(n, min = 3, max = 15),
                type2 = "ltrt",
                family = "weibull")
coef(fit_elife(
   time = x,
   ltrunc = low,
   rtrunc = upp,
   family = "weibull"))
#>    scale    shape 
#> 1.943001 1.579761 
# Simulate left-truncated right-censored data
x <- samp_elife(n = n,
                scale = 2,
                shape = 1.5,
                lower = low <- runif(n),
                upper = upp <- runif(n, min = 3, max = 15),
                type2 = "ltrc",
                family = "gomp")
#note that the return value is a list...
coef(fit_elife(
   time = x$dat,
   ltrunc = low,
   event = !x$rcens,
   family = "gomp"))
#>    scale    shape 
#> 2.336803 1.988250