simdata {rolr}

Simulating Survival Times as Functions of a Single Covariate

Description

simdata is used to simulate survival data from an exponential distribution. When the hazard function is a step function, we assume 3 underlying groups obtained by applying two cutpoints x1 and x2 to the covariate so that group 1 is x < x1, group 2 is x >= x1 and x < x2, and group 3 is x >= x2. The hazard is a function of the covariate x simulated from a uniform distribution from [0, 2]; it can be either a linear function, a step function (with three groups), or a constant (in which case no association exists between the covariate and survival).

Usage

simdata(nn = 300, const = 365, hr = c(1, 2, 3), hr.linear = 3,
  props = c(1/3, 1/3, 1/3), hazard.func = "step", censoring.rate = 0,
  seed = 1)

Arguments

nn

Sample size.

const

A constant that all of the hazard functions will be divided by. The bigger it is, the longer the survival times will be. Default is 365.

hr

A three-element vector representing the hazards for each of the groups 1 to 3 when the hazard.func = "step".

hr.linear

A scalar representing the hazard ratio when the covariate increases by one unit. This is used with hazard.func = "linear".

props

A three-element vector representing the proportions of groups 1 to 3 when hazard.func = "step".

hazard.func

A character that can take either 'step', 'linear', or 'none' to represent a step, linear or no association between the covariate and survival, respectively. When it is "step", the entire set is divided into 3 groups based on the covariate with group proportions specified in the props argument.

censoring.rate

The amount of censoring desired. Default = 0.

seed

The random seed used to generate the data. Default = 1.

Value

A data frame with survival times (times), censoring indicator (status), covariate (x), three groups obtained by cutting the covariate if hazard.func = "step" (x3), and censoring rate (censoring.rate).

Examples

library(rolr)

#simulate survival with a step hazard function
d1=simdata(nn = 150, hr = c(1, 2, 3), props = c(1/3, 1/3, 1/3),
           hazard.func = "step", censoring.rate = 0)
head(d1)

#simulate survival with a linear hazard function
d2=simdata(nn = 150, hr.linear = 2, hazard.func = "linear", censoring.rate = 0)
head(d2)

#simulate survival with no association with the covariate
d3=simdata(nn = 150, hazard.func = "none", censoring.rate = 0)
head(d3)