ggSurv: Survival curve using ggplot2

In bozenne/butils: Various useful functions

Survival curve using ggplot2

Description

Display a non-parametric or semi-parametric survival curve.

Usage

ggSurv(object, ...)

## S3 method for class 'survfit'
ggSurv(object, ...)

## S3 method for class 'coxph'
ggSurv(object, data = NULL, newdata = NULL, confint = FALSE, ...)

## S3 method for class 'data.table'
ggSurv(
  object,
  format = "data.table",
  var.time = "time",
  var.survival = "survival",
  var.ci.inf = NULL,
  var.ci.sup = NULL,
  var.event = NULL,
  var.censor = NULL,
  name.strata = NULL,
  var.strata = NULL,
  plot = TRUE,
  legend.position = "top",
  title = NULL,
  text.size = NULL,
  ylim = NULL,
  line.size = 2,
  confint = FALSE,
  alpha.CIfill = 0.2,
  alpha.CIline = 0.5,
  censoring = FALSE,
  alpha.censoring = 1,
  colour.censoring = rgb(0.2, 0.2, 0.2),
  shape.censoring = 3,
  size.censoring = 2,
  name.censoring = "censoring",
  events = FALSE,
  alpha.events = 1,
  colour.events = rgb(0.2, 0.2, 0.2),
  shape.events = 8,
  size.events = 2,
  name.events = "event",
  ...
)

Arguments

object	a coxph object or a survfit object or a data.table object.
...	arguments to be passed to lower level functions
data	the data that have been used to fit the model.
newdata	dataset containing the covariates conditional to which the survival is displayed.
confint, alpha.CIfill, alpha.CIline	how to display confidence intervals
format	the format used to export the data. Can be data.table or data.frame.
var.time	the name of the column in the data.table object containing the time variable
var.survival	the name of the column in the data.table object containing the survival at each time
var.ci.inf	the name of the column in the data.table object containing the lower bound of the confidence interval
var.ci.sup	the name of the column in the data.table object containing the upper bound of the confidence interval
var.event	the name of the column in the data.table object containing the number of events at each time
var.censor	the name of the column in the data.table object containing the number of censored patients at each time
name.strata	the title for the color and fill legend in the plot.
var.strata	the name of the column in the data.table object containing the strata variable
plot	[logical] should the plot be displayed?
legend.position, title, text.size	annotations on the plot
ylim	y range (survival)
line.size	size of the survival curves
censoring, alpha.censoring, colour.censoring, shape.censoring, size.censoring, name.censoring	how to display censored events
events, alpha.events, colour.events, shape.events, size.events, name.events	how to display non-censored events

Value

A list:

• plot: the ggplot object

• data: the data used to create the ggplot object

Examples

library(lava)
library(survival)
library(data.table)

set.seed(10)
n <- 100
newdata <- data.frame(X1=1)
time <- 0.25

## simulate non proportional hazard using lava
m <- lvm()
regression(m) <- y ~ 1
regression(m) <- s ~ exp(-2*X1+X2)
distribution(m,~X1+X2) <- binomial.lvm()
distribution(m,~cens) <- coxWeibull.lvm(scale=1)
distribution(m,~y) <- coxWeibull.lvm(scale=1,shape=~s)
eventTime(m) <- eventtime ~ min(y=1,cens=0)
d <- as.data.table(lava::sim(m,n))
setkey(d, eventtime)

## Cox (PH)
## no regressor
m.cox <- coxph(Surv(eventtime, status) ~ 1, data = d, y = TRUE, x = TRUE)
res1 <- ggSurv(m.cox)

## regressors
m.cox <- coxph(Surv(eventtime, status) ~ X1, data = d, y = TRUE, x = TRUE)
res1 <- ggSurv(m.cox)
ggSurv(m.cox, newdata = d[,.SD[1], by = X1])
ggSurv(m.cox, newdata = d[,.SD[1], by = X1], confint = TRUE)

## Cox (stratified)
mStrata.cox <- coxph(Surv(eventtime, status) ~ strata(X1), data = d,
                     y = TRUE, x = TRUE)
ggSurv(mStrata.cox, censoring = TRUE, event = TRUE)

mStrata.cox <- coxph(Surv(eventtime, status) ~ strata(X1) + strata(X2), data = d,
                     y = TRUE, x = TRUE)
ggSurv(mStrata.cox, censoring = TRUE, event = TRUE)

## KM
m.KM <- survfit(Surv(eventtime, status) ~ X1, data = d)
res2 <- ggSurv(m.KM)

## combine plot
res1$data[,X1 := as.numeric(factor(strata2, levels = c("X1=0","X1=1")))-1]
res1$data[,strata2 := NULL]
dt.gg <- rbind(cbind(res1$data[original == TRUE & X1==1], model = "cox"),
               cbind(res2$data[original == TRUE & X1==1], model = "KM")
)
dt.gg[,c("X1","original") := NULL]
ggSurv(dt.gg, var.time = "time", var.survival = "survival", var.strata = "model")

bozenne/butils documentation built on Oct. 14, 2023, 6:19 a.m.