The final ggsurvplot would look like this:

library(RTCGA) 
library(RTCGA.clinical)
library(survminer)

BRCAOV.survInfo <- survivalTCGA(BRCA.clinical, OV.clinical, 
                                extract.cols = "admin.disease_code") 
# Survival curves + risk table
fit <- survfit(Surv(times, patient.vital_status)~admin.disease_code,
               data = BRCAOV.survInfo)
surv.curv <- ggsurvplot(fit, pval = TRUE, conf.int = TRUE,
           risk.table = TRUE, legend.labs = c("brca", "ov"))

# n.censor.plot
d <- data.frame(time = fit$time,
                n.censor = fit$n.censor,
                strata = as.factor(rep(names(fit$strata), fit$strata)))

ncensor.plot <- ggplot(d, aes(time, n.censor)) + 
  geom_bar(aes(color = strata, fill = strata), stat = "identity", position = "dodge") +
  theme_classic() +  theme(legend.position = "none") 

# Arrange plots
gridExtra::grid.arrange(surv.curv$plot, surv.curv$table, ncensor.plot, heights = c(0.5, 0.25, 0.25))

Do you have any suggestion about the plot?
Thanks:-)

@prostkop what do You think about that :)?

cens.density = FALSE by defualt :) for me it looks interesting. I am wondering whether bars could be replaced with geom_denisty? What do You think? And You @prostkop?

Marcin Kosinski

Dnia 25.02.2016 o godz. 13:10 Alboukadel KASSAMBARA notifications@github.com napisał(a):

The final ggsurvplot would look like this:

library(RTCGA)
library(RTCGA.clinical)
library(survminer)

BRCAOV.survInfo <- survivalTCGA(BRCA.clinical, OV.clinical,
extract.cols = "admin.disease_code")

Survival curves + risk table

fit <- survfit(Surv(times, patient.vital_status)~admin.disease_code,
data = BRCAOV.survInfo)
surv.curv <- ggsurvplot(fit, pval = TRUE, conf.int = TRUE,
risk.table = TRUE, legend.labs = c("brca", "ov"))

n.censor.plot

d <- data.frame(time = fit$time,
n.censor = fit$n.censor,
strata = as.factor(rep(names(fit$strata), fit$strata)))

ncensor.plot <- ggplot(d, aes(time, n.censor)) +
geom_bar(aes(color = strata, fill = strata), stat = "identity", position = "dodge") +
theme_classic() + theme(legend.position = "none")

Arrange plots

gridExtra::grid.arrange(surv.curv$plot, surv.curv$table, ncensor.plot, heights = c(0.5, 0.25, 0.25))

Do you have any suggestion about the plot?
Thanks:-)

—
Reply to this email directly or view it on GitHub.

I think this 3rd plot should have the same width of axis as the first 2 containing survival curves and risk table

Definitely, last plot should be in line with others. In addition, on the 3rd plot number of censoring is hardly visible, because of a peak in the beginning of the study. Maybe, some "algorithm" should be applied to mark "outlying" bars and decrease their heights. But essentially, it is very nice plot :)

And in my opinion colours of the last plot should vary from the colours of the survival curves, so to avoid confusion :)

Yes, I completely agree with you!! This was a quick arrangement of the plots for discussion... I'll work in the script to improve the final output!

Thanks :)!

Hey, have you tried working on that :)?

Hi Marcin,

I think that plotting the number of censor events, at each time point, can be really unreadable in a situation where you have a large cohort of patients with multiple groups.

For example (option 1):

library(survival)
library(survminer)
fit <- survfit( Surv(time, status) ~ rx + adhere,
                 data = colon )
surv.curv <- ggsurvplot(fit, risk.table = TRUE, risk.table.col = "strata",
                      break.time.by = 500)
surv.curv

The plot of censor events:

# ncensor data
d <- data.frame(time = fit$time,
                n.censor = fit$n.censor,
                strata = as.factor(rep(names(fit$strata), fit$strata)))

ggplot(d, aes(time, n.censor)) + 
  geom_bar(aes(color = strata, fill = strata), stat = "identity", position = "dodge") +
  theme_bw() +
  ggplot2::theme(legend.position = "none")

However, we can plot the number of censor event at specific time points to make the plot more readable (option 2):

times <- ggplot_build(surv.curv$plot)$panel$ranges[[1]]$x.major_source

# n.censor.plot
fit_summary <- summary(fit, times = times, extend = TRUE)
d <- data.frame(time = as.factor(fit_summary$time),
                n.censor = fit_summary$n.censor,
                strata = factor(fit_summary$strata)
                  )

# bar plot of censor events
ggplot(d, aes(time, n.censor)) + 
  geom_bar(aes(color = strata, fill = strata), stat = "identity", position = "dodge") +
  theme_classic() +
  ggplot2::theme(legend.position = "none")

# line plot of censor events
d$time <- fit_summary$time
ggplot(d, aes(time, n.censor)) + 
  geom_line(aes(color = strata))+
  theme_classic() +
  ggplot2::theme(legend.position = "none")

Bar plot:

Line plot:

The final plot would look like as follow:

# censor plot
cens.plot <- ggplot(d, aes(time, n.censor)) + 
  geom_line(aes(color = strata))+
  theme_classic() +
  ggplot2::theme(legend.position = "none")+
  coord_cartesian(xlim = c(0, max(fit$time))) +
  ggplot2::scale_x_continuous(breaks = times)+
  theme(legend.key.height = NULL, legend.key.width = NULL,
       legend.key = element_rect(colour = NA, fill = NA),
       legend.text = element_text(colour = NA),
       legend.title = element_text(colour = NA)) +
  guides(color = FALSE)

# Arrange plots
x <- list(ncensor = cens.plot, table = surv.curv$table, plot = surv.curv$plot)
plots <- rev(x)
grobs <- widths <- list()
for (i in 1:length(plots)) {
  grobs[[i]] <- ggplotGrob(plots[[i]])
  widths[[i]] <- grobs[[i]]$widths[2:5]
}
maxwidth <- do.call(grid::unit.pmax, widths)
for (i in 1:length(grobs)) {
  grobs[[i]]$widths[2:5] <- as.list(maxwidth)
}
do.call(gridExtra::grid.arrange, c(grobs, nrow = 3, heights = list(c(0.5, 0.25, 0.25))))

What do you think about option 2?

I think line plot and barplot are great ideas :) There could be a parameter censor.type = c('line', 'barplot') ? Great job!

@prostkop do you have any suggestions?

Ok! I'm adding these options in ggsurvplot function:-)!

Great! Thanks :)

2016-03-23 17:26 GMT+01:00 Alboukadel KASSAMBARA notifications@github.com:

Ok! I'm adding these options in ggsurvplot function:-)!

—
You are receiving this because you authored the thread.
Reply to this email directly or view it on GitHub
#18 (comment)

Sorry for a litle delay. Indeed, plot looks really tidy and nice. I don't have time to run your example, but do you choose some specific time points for n.cesoring plot and between these points linear interpolations is used ? How about using all time points with censoring ? Probably, plot will be more irregular, but maybe more reliable ?

@kassambara can we close this issue? Is that feature avaiable?

Hi!!
I will finalize this in v0.2.2. I planed to submit v0.2.1 to CRAN tomorrow. I updated the Readme.Rmd and built a staticdocs (http://www.sthda.com/english/rpkgs/survminer/) so that users can check out the documentation and usage examples of each function in survminer package.

Wow, this looks great. Can you explain how this was created and is it the
same tool as for ggplo2 documentation?

1. I was about to suggest you to use readthedocs for
   documentation - what do you think about that?
   1a) I was also thinking about a website for survminer package. It can be
   hosted on github pages (or wherever you have bought hosting). It can be
   done in jekyll (using jekyll-knitr (motivation, themes to choose from). I have experience with jekyll since I have
   created 2 websites using that tool

• rtcga.github.io/RTCGA/
• http://r-addict.com

2. Thanks for adding me to authors. This is like an award for me, since
   this package is really great (and even though survMisc is available on
   CRAN again) . I was showing this package during "R workshops for young
   biotechnologists" that I have lead today and yesterday in Warsaw.
   Moreover, I would like to contribute more with this package.
   2a) I think to have complete cox analysis we could also add something like
   ggcoxNAME for full (simple) coxph model diagnostics like plots of
   residuals vs linear combinations of explanatory variables.

3. I know that you were planning to blog about survminer soon. I also was
   thinking about additional blog post. More posts, more users :) Maybe it
   would be good not to post our both posts in the same week?

I know I've mentioned a lot of topics. You can split answers to separate
issues.

PS. Have you thought about attending eRum :) http://erum.ue.poznan.pl/?

2016-04-22 0:04 GMT+02:00 Alboukadel KASSAMBARA notifications@github.com:

Hi!!
I will finalize this in v0.2.2. I planed to submit v0.2.1 to CRAN
tomorrow. I updated the Readme.Rmd and built a staticdocs (
http://www.sthda.com/english/rpkgs/survminer/) so that users can check
out the documentation and usage examples of each function in survminer
package.

—
You are receiving this because you authored the thread.
Reply to this email directly or view it on GitHub
#18 (comment)

I used the R package staticdocs, developed by Hadley Wickham, to build survminer documentation website. ggplot2 online documentation has been built using the same package.

The job is simply done using the following R code:

staticdocs::build_site(pkg =  "local/path/to/survminer", site_path = "path/to/site/directory")

1. It would be nice to build a website for survminer. I have not tried yet jekyll but I have to:-)! I think that hosting on github is the best option.

2. You have made significant contributions to the package. Thank you for your great work!!

3. I agree with you (see Blog post about the short history of how survival plots were improved between survminer ver 0.1 and ver 0.2? #34 )

PS. I'm not sure to attend eRum

Best,
A.

I think this is still an interesting feature and we could think about it in ver 0.2.3 ? :)

New argument 'ncensor.plot' added to ggsurvplot(). A logical value. If TRUE, the number of censored subjects at time t is plotted. Default is FALSE.

require("survival")
fit<- survfit(Surv(time, status) ~ sex, data = lung)
ggsurvplot(fit, risk.table = TRUE, ncensor.plot = TRUE, risk.table.height = 0.35)

Marvelous :) Keep going @kassambara
Your on a programming spree :D