Confidence Intervals and Confidence Bands for the Predicted Absolute Risk (Cumulative Incidence Function)
Confidence intervals and confidence Bands for the predicted absolute risk (cumulative incidence function).
## S3 method for class 'predictCSC' confint( object, parm = NULL, level = 0.95, n.sim = 10000, absRisk.transform = "loglog", seed = NA, ... )
object |
A |
parm |
not used. For compatibility with the generic method. |
level |
[numeric, 0-1] Level of confidence. |
n.sim |
[integer, >0] the number of simulations used to compute the quantiles for the confidence bands. |
absRisk.transform |
[character] the transformation used to improve coverage
of the confidence intervals for the predicted absolute risk in small samples.
Can be |
seed |
[integer, >0] seed number set before performing simulations for the confidence bands. If not given or NA no seed is set. |
... |
not used. |
The confidence bands and confidence intervals are automatically restricted to the interval [0;1].
Brice Ozenne
library(survival)
library(prodlim)
#### generate data ####
set.seed(10)
d <- sampleData(100)
#### estimate a stratified CSC model ###
fit <- CSC(Hist(time,event)~ X1 + strata(X2) + X6, data=d)
#### compute individual specific risks
fit.pred <- predict(fit, newdata=d[1:3], times=c(3,8), cause = 1,
se = TRUE, iid = TRUE, band = TRUE)
fit.pred
## check confidence intervals
newse <- fit.pred$absRisk.se/(-fit.pred$absRisk*log(fit.pred$absRisk))
cbind(lower = as.double(exp(-exp(log(-log(fit.pred$absRisk)) + 1.96 * newse))),
upper = as.double(exp(-exp(log(-log(fit.pred$absRisk)) - 1.96 * newse)))
)
#### compute confidence intervals without transformation
confint(fit.pred, absRisk.transform = "none")
cbind(lower = as.double(fit.pred$absRisk - 1.96 * fit.pred$absRisk.se),
upper = as.double(fit.pred$absRisk + 1.96 * fit.pred$absRisk.se)
)Please choose more modern alternatives, such as Google Chrome or Mozilla Firefox.