The goal of crrf is to enable fitting of the Firth-penalized Fine-Gray model for competing risks regression. The Firth penalty allows for parameter estimation in the case of monotone likelihood. This work was inspired by a SAS macro by Kohl et al., 2016 (https://doi.org/10.1016/j.cmpb.2014.11.009).
You can install the development version of crrf from GitHub with:
# install.packages("devtools")
devtools::install_github("annaSeffernick/crrf")This is a basic example which shows you how to solve a common problem:
library(crrf)
# Generate Data
set.seed(12345)
n=100
tm=rexp(n)
ev=factor(rbinom(n,2,0.3))
x=rnorm(n)
y=rbinom(n,1,0.5)
u=runif(n)
dset=cbind.data.frame(tm=tm,ev=ev,x=x,y=y,u=u)
# fit the model
crrf.fit=crrf(Surv(tm,ev)~x+y+u,etype="1",dset,firth=TRUE,CI=TRUE)
crrf.fit$CI.tbl
#> ml.beta lower upper
#> x 0.1656148 -0.16418727 0.4948082
#> y 0.6365446 0.02979408 1.2726049
#> u 0.1264342 -0.94058592 1.2238442