simfms: Simulation of clustered multi-state data

In simfms: Simulation of clustered multi-state data

Description

Random generation of survival data with a clustered and multi-state structure.

The simulation procedure is based on a copula model for each competing events block, allowing to specify the marginal distributions of time variables.

The effect of simulated frailties and covariates can be added in a proportional hazards way.

Usage

simfms(nsim = NULL, tmat = NULL, clock = "forward", 
       frailty = list(dist="gamma", par= .5, type="shared"), 
       nclus = NULL, csize = NULL, 
       covs = NULL, beta = NULL, 
       marg = list(dist = "weibull", lambda = 1, rho = 1), 
       cens = list(dist = "weibull", lambda = 1, rho = 1, admin = 72), 
       copula = list(name = "clayton", par = 1),
       format = "long")

Arguments

nsim	The number of subjects to simulate
tmat	The transitions matrix
clock	The approach to be used for tractation of baseline hazards and generation of times: reset generates times at each step ranging from 0 to infinity and adds them to those of previous transitions forward generates times at each step, conditional on the time of entering the present state
frailty	The frailty term specifications. A list with components: dist: the name of the frailty distribution (only gamma implemented at now) par: the frailty parameter(s) value type: the type of frailty shared: iid and shared by the times of transitions of any type, for all subjects in a group iid: iid and shared by the times of transitions of the same type, for all subjects in a group (possible different frailty parameters for different transition types) nested: id and correlated by means of nested frailties (possible different frailty parameters for different transition types)
nclus	The number of clusters to simulate
csize	The size(s) of cluster
covs	The covariates to simulate. A list with components nameOfCovariate = simulationfunction
beta	The regression coefficients for covariates. A list of the same length as covs, with elements nameOfCovariate, vectors of the same length as the number of transitions max(tmat, na.rm=TRUE)
marg	The marginal baseline hazards. A list with components dist: the name of the baseline hazard distributions (one value for all transitions) eachpar: each baseline parameter (either one value or as many as the number of transitions in codetmat)
cens	The censoring time distributions. A list with components dist: the name of the censoring distributions (one value for all transitions) eachpar: each censoring distribution parameter (either one value or as many as the number of possible starting states in tmat, possibly with states' names)
copula	The copula model. A list with components name: the name of the copula (only clayton available at now) par: the copula parameter
format	the data format, either long or wide

Value

A data.frame containing the simulated data, with columns

Author(s)

Federico Rotolo <[e-mail]>

References

Rotolo, F, Legrand, Catherine, Van Keilegom, I (2012) Simulation of clustered multi-state survival data based on a copula model. Submitted

See Also

trans.cancer

Examples

################################################################################
### - Cancer reduced multi-state structure - ###################################
################################################################################
trans.cancer.reduced()
simfms(nsim  = NULL,
       tmat  = trans.cancer.reduced(),
       clock = "forward",
       frailty = list(dist="gamma", par= .5),
       nclus = 5,
       csize = 2,
       covs = list(age=function(x) rnorm(x, mean=60, sd=7),
                   treat=function(x) rbinom(x, 1, .5)),
       beta = list(age=rep(.02,5), treat=rep(2,5)),
       marg  = list(dist="weibull", lambda=1, rho=1), 
       cens  = list(dist="weibull", lambda=1, rho=1, admin= 72),
       copula= list(name="clayton", par= 1))


################################################################################
### - Cancer standard multi-state structure - ##################################
################################################################################
trans.cancer()
simfms(nsim  = NULL,
       tmat  = trans.cancer(),
       clock = "forward",
       frailty = list(dist="gamma", par= .5),
       nclus = 5,
       csize = 2,
       covs = list(age=function(x) rnorm(x, mean=60, sd=7),
                   treat=function(x) rbinom(x, 1, .5)),
       beta = list(age=rep(.02, 8), treat=rep(2, 8)),
       marg  = list(dist="weibull", lambda=1, rho=1), 
       cens  = list(dist="weibull", lambda=1, rho=1,  admin= 72),
       copula= list(name="clayton", par= 1))


################################################################################
### - Cancer extended multi-state structure - ##################################
################################################################################
trans.cancer.extended()
simfms(nsim  = NULL,
       tmat  = trans.cancer.extended(),
       clock = "forward",
       frailty = list(dist="gamma", par= .5),
       nclus = 5,
       csize = 2,
       covs = list(age=function(x) rnorm(x, mean=60, sd=7),
                   treat=function(x) rbinom(x, 1, .5)),
       beta = list(age=rep(.02, 9), treat=rep(2, 9)),
       marg  = list(dist="weibull", lambda=1, rho=1), 
       cens  = list(dist="weibull", lambda=1, rho=1, admin= 72),
       copula= list(name="clayton", par= 1))

simfms documentation built on May 2, 2019, 6:50 p.m.