survfit_phregr function

Survival Curve for Proportional Hazards Regression Models

Obtains the predicted survivor function for a proportional hazards regression model.

survfit_phregr(
  fit_phregr,
  newdata,
  sefit = TRUE,
  conftype = "log-log",
  conflev = 0.95
)

Arguments

• fit_phregr: The output from the phregr call.
• newdata: A data frame with the same variable names as those that appear in the phregr call. For right-censored data, one curve is produced per row to represent a cohort whose covariates correspond to the values in newdata. For counting-process data, one curve is produced per id in newdata to present the survival curve along the path of time-dependent covariates at the observed event times in the data used to fit phregr.
• sefit: Whether to compute the standard error of the survival estimates.
• conftype: The type of the confidence interval. One of "none", "plain", "log", "log-log" (the default), or "arcsin". The arcsin option bases the intervals on asin(sqrt(surv)).
• conflev: The level of the two-sided confidence interval for the survival probabilities. Defaults to 0.95.

Returns

A data frame with the following variables:

• id: The id of the subject for counting-process data with time-dependent covariates.
• time: The observed times in the data used to fit phregr.
• nrisk: The number of patients at risk at the time point in the data used to fit phregr.
• nevent: The number of patients having event at the time point in the data used to fit phregr.
• cumhaz: The cumulative hazard at the time point.
• surv: The estimated survival probability at the time point.
• sesurv: The standard error of the estimated survival probability.
• lower: The lower confidence limit for survival probability.
• upper: The upper confidence limit for survival probability.
• conflev: The level of the two-sided confidence interval.
• conftype: The type of the confidence interval.
• covariates: The values of covariates based on newdata.
• stratum: The stratum of the subject.

Details

If newdata is not provided and there is no covariate, survival curves based on the basehaz data frame will be produced.

Examples

library(dplyr)

# Example 1 with right-censored data
fit1 <- phregr(data = rawdata %>% filter(iterationNumber == 1) %>%
                 mutate(treat = 1*(treatmentGroup == 1)),
               stratum = "stratum",
               time = "timeUnderObservation", event = "event",
               covariates = "treat")

surv1 <- survfit_phregr(fit1,
                        newdata = data.frame(
                          stratum = as.integer(c(1,1,2,2)),
                          treat = c(1,0,1,0)))

# Example 2 with counting process data and robust variance estimate
fit2 <- phregr(data = heart %>% mutate(rx = as.numeric(transplant) - 1),
               time = "start", time2 = "stop", event = "event",
               covariates = c("rx", "age"), id = "id", robust = TRUE)

surv2 <- survfit_phregr(fit2,
                        newdata = data.frame(
                          id = c(4,4,11,11),
                          age = c(-7.737,-7.737,-0.019,-0.019),
                          start = c(0,36,0,26),
                          stop = c(36,39,26,153),
                          rx = c(0,1,0,1)))

References

Terry M. Therneau and Patricia M. Grambsch. Modeling Survival Data: Extending the Cox Model. Springer-Verlag, 2000.

Author(s)

Kaifeng Lu, kaifenglu@gmail.com