fastCRRC
implements a fast and efficient clustered competing risks
(CCR) regression model for large correlated clustered data. This
regression model assesses the marginal effects of covariates on the
respective cumulative incidence functions while accommodating the
correlation induced due to clustering. Parameter estimation utilizes the
‘forward-backward scan’ algorithm to drastically reduce the computation
time for large clustered competing risk datasets. Variance estimation is
efficiently handled using a clustered bootstrap approach.
Let
The marginal cumulative incidence function conditional on the covariates
is defined as
where
Vignettes is available here
You can install the development version of fastCRRC
from
GitHub with:
# install.packages("devtools")
devtools::install_github("edemprd/fastCRRC")
This is a basic example which shows you how to simulate data and fit the CCR model using the package.
library(fastCRRC)
set.seed(1)
dat <- simCR(n = 10, al = .3, ga = .6, clsize = 2:10, summary = TRUE)
#> Call:
#> simCR(n = 10, al = 0.3, ga = 0.6, clsize = 2:10, summary = TRUE)
#>
#> Number of clusters generated: 10
#> Average number of observations per cluster: 7.9
#> Overall observed censoring rate: 0.722
#>
#> Before censoring:
#> Proportion of event 1: 0.772
#> Proportion of event 2: 0.228
#> Before censoring:
#> Proportion of event 1: 0.253
#> Proportion of event 2: 0.025
#> Average proportion of event 1 per cluster: 0.235
#> Average proportion of event 2 per cluster: 0.029
fit1 <- fastCrrC(Surv(time, status) ~ Z.1 + Z.2, data = dat, B = 100, fitter = "fastCrr")
summary(fit1)
#> Call:
#> fastCrrC(formula = Surv(time, status) ~ Z.1 + Z.2, data = dat,
#> fitter = "fastCrr", B = 100)
#>
#> fastCrrC Estimator
#> estimate std.Error z.value p.value
#> Z.1 0.4390 0.2709 1.621 0.1050
#> Z.2 0.0181 1.2406 0.015 0.9884
-
Zhou, B., Fine, J., Latouche, A., & Labopin, M. (2012). Competing risks regression for clustered data. Biostatistics, 13(3), 371-383.
-
Kawaguchi, E. S., Shen, J. I., Suchard, M. A., & Li, G. (2021). Scalable algorithms for large competing risks data. Journal of Computational and Graphical Statistics, 30(3), 685-693.