Adding Line Colour Changes Line Position

[DarioS]

I firstly used the command

survivalPlot <- autoplot(survivalFit, conf.int = FALSE)

and got a sensible looking plot.

Then, I added a line colour and the placement of lines changed

autoplot(survivalFit, conf.int = FALSE, surv.colour = "red")

The legend also disappeared.

[averissimo]

You want to control the colors on each curve? try this:

p1 <- autoplot(survival_obj)
p1 <- p1 + scale_colour_manual(values = c('indianred2','seagreen'))

I was having the same problem and when debugging found that it uses geom_step function and this is how to control the colors :)

Despite all this, I think the implementation on ggfortify should be more clear on that option or corrected.

Cheers

[sinhrks]

@averissimo thx for the reply. It is correct that we should specify multi colors like ggplot2 does.
Maybe better to update the doc.

@DarioS Sorry to late reply, I missed the notification.

[DarioS]

I think that it would be better to provide abstraction of this to the end-user, as the function currently tries to do unsuccessfully. The underlying ggplot2 framework should be hidden away where possible.

[sinhrks]

OK, will consider the fix.

[lselzer]

@DarioS I'm guessing that you want to make all lines red regardless of which group they are. Am I correct?

[DarioS]

Actually, it would be desirable to specify one colour per factor level. I only used "red" because the surv.colour option doesn't allow a vector of values.

> library(survival)
> autoplot(survfit(Surv(time, status) ~ sex, data = lung), surv.colour = c("red", "blue"))
Error: Aesthetics must be either length 1 or the same as the data (208): colour

Note that the number of samples in the error message is wrong.

> dim(lung)
[1] 228  10

[lselzer]

I will take a look tomorrow. I think the issue is with the way that geom_factory() works. Am I right @sinhrks ? I likely fix would for geom_factory() to check if the column exists in the data if not, check if it is a valid colour.

[terrytangyuan]

A pull request for fixing this would be appreciated! @lselzer @DarioS @averissimo