riskRegression (version 1.4.3)

predictCox: Fast computation of survival probabilities, hazards and cumulative hazards from Cox regression models

Description

Fast routine to get baseline hazards and subject specific hazards as well as survival probabilities from a survival::coxph or rms::cph object

Usage

predictCox(object, newdata = NULL, times, centered = TRUE,
  type = c("cumhazard", "survival"), keep.strata = TRUE,
  keep.times = TRUE, keep.newdata = FALSE, se = FALSE, band = FALSE,
  iid = FALSE, average.iid = FALSE, nSim.band = 10000,
  conf.level = 0.95, logTransform = TRUE, store.iid = "full")

Arguments

object

The fitted Cox regression model object either obtained with coxph (survival package) or cph (rms package).

newdata

A data.frame or data.table containing the values of the predictor variables defining subject specific predictions. Should have the same structure as the data set used to fit the object.

times

Time points at which to evaluate the predictions.

centered

Logical. If TRUE return prediction at the mean values of the covariates fit$mean, if FALSE return a prediction for all covariates equal to zero. in the linear predictor. Will be ignored if argument newdata is used.

type

the type of predicted value. Choices are

• "hazard" the baseline hazard function when argument newdata is not used and the hazard function when argument newdata is used.

• "cumhazard" the cumulative baseline hazard function when argument newdata is not used and the cumulative hazard function when argument newdata is used.

• "survival" the survival baseline hazard function when argument newdata is not used and the cumulative hazard function when argument newdata is used.

Several choices can be combined in a vector of strings that match (no matter the case) strings "hazard","cumhazard", "survival".

keep.strata

Logical. If TRUE add the (newdata) strata to the output. Only if there any.

keep.times

Logical. If TRUE add the evaluation times to the output.

keep.newdata

Logical. If TRUE add the value of the covariates used to make the prediction in the output list.

se

Logical. If TRUE add the standard error to the output.

band

Logical. If TRUE add the confidence band to the output.

iid

Logical. If TRUE add the influence function to the output.

average.iid

Logical. If TRUE add the average of the influence function over newdata to the output.

nSim.band

the number of simulations used to compute the quantiles for the confidence bands..

conf.level

Level of confidence.

logTransform

Should the confidence intervals/bands be computed on the log (hazard) and log(-log) (survival) scale and be backtransformed. Otherwise they are computed on the original scale and truncated (if necessary).

store.iid

Implementation used to estimate the influence function and the standard error. Can be "full" or "minimal". See the details section of calcSeCox.

...

arguments to be passed to the function iidCox.

Value

A list with some or all of the following elements:

• times: the time points at which the other elements are evaluated.

• hazard: When argument newdata is not used the baseline hazard function, otherwise the predicted hazard function.

• cumhazard: When argument newdata is not used the cumulative baseline hazard function, otherwise the predicted cumulative hazard function.

• cumhazard.se: The standard errors of the predicted cumulative hazard function.

• survival: When argument newdata is not used the survival probabilities corresponding to the baseline hazard, otherwise the predicted survival probabilities.

• survival.se: The standard errors of the predicted survival.

• strata: The strata variable.

Details

Not working with time varying predictor variables or delayed entry. The centered argument enables us to reproduce the results obtained with the basehaz function from the survival package.

When setting logTransform to TRUE, the standard error that is returned is before back-transformation to the original scale.

Examples

# NOT RUN {
library(survival)

set.seed(10)
d <- sampleData(40,outcome="survival")
nd <- sampleData(4,outcome="survival")
d$time <- round(d$time,1)
fit <- coxph(Surv(time,event)~X1 + strata(X2) + X6,
             data=d, ties="breslow", x = TRUE, y = TRUE)
# table(duplicated(d$time))

predictCox(fit)
predictCox(fit,centered=FALSE,type="hazard")
predictCox(fit,centered=TRUE,type="hazard")
predictCox(fit, newdata=nd, times=c(3,8),se=TRUE)
predictCox(fit, newdata=nd, times=c(3,8),se=TRUE, logTransform = TRUE)
predictCox(fit, newdata=nd, times=c(3,8),se=TRUE, store.iid = "minimal")
predictCox(fit, newdata=nd, times = 5,iid=TRUE)

cbind(survival::basehaz(fit),predictCox(fit,type="cumhazard")$cumhazard)

# one strata variable
fitS <- coxph(Surv(time,event)~strata(X1)+X2,
              data=d, ties="breslow", x = TRUE, y = TRUE)

predictCox(fitS)
predictCox(fitS, newdata=nd, times = 1)

# two strata variables
set.seed(1)
d$U=sample(letters[1:5],replace=TRUE,size=NROW(d))
d$V=sample(letters[4:10],replace=TRUE,size=NROW(d))
nd$U=sample(letters[1:5],replace=TRUE,size=NROW(nd))
nd$V=sample(letters[4:10],replace=TRUE,size=NROW(nd))
fit2S <- coxph(Surv(time,event)~X1+strata(U)+strata(V)+X2,
              data=d, ties="breslow", x = TRUE, y = TRUE)

cbind(survival::basehaz(fit2S),predictCox(fit2S,type="cumhazard")$cumhazard)
predictCox(fit2S)
predictCox(fitS, newdata=nd, times = 3)


# }