node_cox {simDAG} | R Documentation |
Simulate a Node Using Cox-Regression
Description
Data from the parents is used to generate the node using cox-regression using the method of Bender et al. (2005).
Usage
node_cox(data, parents, formula=NULL, betas, surv_dist, lambda, gamma,
cens_dist, cens_args, name)
Arguments
data |
A |
parents |
A character vector specifying the names of the parents that this particular child node has. If non-linear combinations or interaction effects should be included, the user may specify the |
formula |
An optional |
betas |
A numeric vector with length equal to |
surv_dist |
A single character specifying the distribution that should be used when generating the survival times. Can be either |
lambda |
A single number used as parameter defined by |
gamma |
A single number used as parameter defined by |
cens_dist |
A single character naming the distribution function that should be used to generate the censoring times. For example, |
cens_args |
A list of named arguments which will be passed to the function specified by the |
name |
A single character string specifying the name of the node. |
Details
The survival times are generated according to the cox proportional-hazards regression model as defined by the user. How exactly the data-generation works is described in detail in Bender et al. (2005). To also include censoring, this function allows the user to supply a function that generates random censoring times. If the censoring time is smaller than the generated survival time, the individual is considered censored.
Unlike the other node
type functions, this function adds two columns to the resulting dataset instead of one. The first column is called paste0(name, "_event")
and is a logical variable, where TRUE
indicates that the event has happened and FALSE
indicates right-censoring. The second column is named paste0(name, "_time")
and includes the survival or censoring time corresponding to the previously mentioned event indicator. This is the standard format for right-censored time-to-event data without time-varying covariates.
To simulate more complex time-to-event data, the user may need to use the sim_discrete_time
function instead.
Value
Returns a data.table
of length nrow(data)
containing two columns. Both starting with the nodes name
and ending with _event
and _time
. The first is a logical vector, the second a numeric one.
Author(s)
Robin Denz
References
Bender R, Augustin T, Blettner M. Generating survival times to simulate Cox proportional hazards models. Statistics in Medicine. 2005; 24 (11): 1713-1723.
Examples
library(simDAG)
set.seed(3454)
# define DAG
dag <- empty_dag() +
node("age", type="rnorm", mean=50, sd=4) +
node("sex", type="rbernoulli", p=0.5) +
node("death", type="cox", parents=c("sex", "age"), betas=c(1.1, 0.4),
surv_dist="weibull", lambda=1.1, gamma=0.7, cens_dist="runif",
cens_args=list(min=0, max=1))
sim_dat <- sim_from_dag(dag=dag, n_sim=1000)