survPHplot: Plot to assess non-proportionality

Description

Plot of log(time) versus -log(-log(survival)) to assess non-proportionality. A constant distance between curves suggest proportionality.

Usage

survPHplot(formula, data, subset, contrasts, weights, col = 1:5,
           lty = 1:5, pch = 19, xlab = "Time (log scale)",
           ylab = "-log(-log(Survival))", log = "x",
           legend.args = list(), ...)

Value

Primary purpose is for plotting (side effect). The return value is initial plot.

Arguments

formula: either (i) formula with a Surv object on the left-hand-side and stratification covariates on the right-hand-side, or (ii) a survfit object
data: data argument passed to survfit
subset: subset argument passed to survfit
contrasts: contrasts argument passed to survfit
weights: weights argument passed to survfit
col: colours of the curves passed to lines
lty: line type of the curves passed to lines
pch: pch for the curves passed to points
xlab: xlab graphics argument passed to plot.default
ylab: ylab graphics argument passed to plot.default
log: log graphics argument passed to plot.default
legend.args: list of arguments passed to legend. These arguments update the base arguments, which are list(x="topright",legend=names(survfit$strata),col=col,lty=lty,pch=pch)
...: Other arguments passed to plot.default

Details

The default plot is to use straight lines between the transformed survival values for each strata, rather than using steps.

Examples

Run this code

survPHplot(Surv(surv_mm/12, status == "Dead: cancer") ~ year8594,
           data=colon, subset=(stage=="Localised"),
           legend.args=list(bty="n"))

Run the code above in your browser using DataLab