sim.cox: Simulation of output from Cox model.

Description

Simulates data that looks like fit from Cox model. Censor data automatically for highest value of the event times by using cumulative hazard.

Usage

# S3 method for cox
sim(cox, n, data = NULL, cens = NULL, rrc = NULL,
  entry = NULL, ...)

Arguments

cox

output form coxph or cox.aalen model fitting cox model.

number of simulations.

data

to extract covariates for simulations (draws from observed covariates).

cens

specifies censoring model, if "is.matrix" then uses cumulative hazard given, if "is.scalar" then uses rate for exponential, and if not given then takes average rate of in simulated data from cox model.

rrc

possible vector of relative risk for cox-type censoring.

entry

delayed entry variable for simulation.

...

arguments for pc.hazard, for example entry-time

Examples

Run this code

# NOT RUN {
data(TRACE)

cox <-  coxph(Surv(time,status==9)~vf+chf+wmi,data=TRACE)
sim1 <- sim.cox(cox,1000,data=TRACE)
cc <- coxph(Surv(time,status)~vf+chf+wmi,data=sim1)
cbind(cox$coef,cc$coef)

cor(sim1[,c("vf","chf","wmi")])
cor(TRACE[,c("vf","chf","wmi")])

cox <-  cox.aalen(Surv(time, status==9) ~ prop(vf)+prop(chf)+prop(wmi),TRACE,robust=0)
sim2 <- sim.cox(cox,1000,data=TRACE)
cc <-  cox.aalen(Surv(time, status)~prop(vf)+prop(chf)+prop(wmi),data=sim2,robust=0)
###
plot(cox)
lines(cc$cum,type="s",col=2)
cbind(cox$gamma,cc$gamma)

# }
# NOT RUN {
### do not test to avoid dependence on mets
library(mets)
cox <-  phreg(Surv(time, status==9)~vf+chf+wmi,data=TRACE)
sim3 <- sim.cox(cox,1000,data=TRACE)
cc <-  phreg(Surv(time, status)~vf+chf+wmi,data=sim3)
cbind(cox$coef,cc$coef)
basehazplot.phreg(cox,se=TRUE)
lines(cc$cumhaz,col=2)

cox <-  phreg(Surv(time,status==9)~strata(chf)+vf+wmi,data=TRACE)
sim3 <- sim.cox(cox,1000,data=TRACE)
cc <-   phreg(Surv(time, status)~strata(chf)+vf+wmi,data=sim3)
cbind(cox$coef,cc$coef)
basehazplot.phreg(cox)
basehazplot.phreg(cc,add=TRUE)
# }
# NOT RUN {
# }

Run the code above in your browser using DataCamp Workspace