r - ggplot2 绘制 Kaplan-Meier 图时在生存数据框中找到空行

标签 r graphics graph plot ggplot2

我正在尝试使用ggplot2绘制一些Kaplan-Meier曲线,代码位于:https://github.com/kmiddleton/rexamples/blob/master/qplot_survival.R

我在不同的数据库中使用这段出色的代码获得了良好的结果。但是,在这种情况下,它给了我以下错误......就好像我的数据框中有空行:

错误 en if (nrow(layer_data) == 0) return() :参数长度为零。

之前关于此类错误的问题似乎对我来说没有用,因为我的情况下数据和函数的类型并不相同。

我对使用 R 进行统计分析相当陌生,而且我没有编程背景,所以我认为这一定是我的数据中的“愚蠢错误”,但我找不到它在哪里......看起来确实如此ggplot2 无法找到要绘制的行。请问,您能以任何方式帮助我,提供线索、建议等吗?

这是我的数据和按顺序使用的代码,为控制台做好了准备-我在knitr脚本中尝试了它-。最后,我发布了我的 sessionInfo:

library(splines)
library(survival)
library(abind)
library(ggplot2)
library(grid)

我创建了一个名为acbi30的数据框(真实数据):

mort28day <- c(1,0,1,0,0,0,0,1,0,0,0,1,1,0,1,0,0,1,0,1,1,1,1,0,1,1,1,0,0,1)
daysurv <- c(4,29,24,29,29,29,29,19,29,29,29,3,9,29,15,29,29,11,29,5,13,20,22,29,16,21,9,29,29,15)
levo <- c(0,0,0,0,1,1,1,0,0,0,0,0,0,0,0,0,0,1,0,0,1,0,0,0,0,0,0,0,0,0)
acbi30 <- data.frame(mort28day, daysurv, levo)
save(acbi30, file="acbi30.rda")
acbi30

然后,我粘贴以下命令以使用 ggplot2 创建函数:

t.Surv <- Surv(acbi30$daysurv, acbi30$mort28day)
t.survfit <- survfit(t.Surv~1, data=acbi30)


#define custom function to create a survival data.frame#
createSurvivalFrame <- function(f.survfit){

#initialise frame variable#
f.frame <- NULL

#check if more then one strata#
if(length(names(f.survfit$strata)) == 0){

#create data.frame with data from survfit#
f.frame <- data.frame(time=f.survfit$time, n.risk=f.survfit$n.risk, n.event=f.survfit$n.event, n.censor = f.survfit
$n.censor, surv=f.survfit$surv, upper=f.survfit$upper, lower=f.survfit$lower)
#create first two rows (start at 1)#
f.start <- data.frame(time=c(0, f.frame$time[1]), n.risk=c(f.survfit$n, f.survfit$n), n.event=c(0,0),
n.censor=c(0,0), surv=c(1,1), upper=c(1,1), lower=c(1,1))
#add first row to dataset#
f.frame <- rbind(f.start, f.frame)
#remove temporary data#
rm(f.start)
}
else {
#create vector for strata identification#
f.strata <- NULL
for(f.i in 1:length(f.survfit$strata)){
#add vector for one strata according to number of rows of strata#
f.strata <- c(f.strata, rep(names(f.survfit$strata)[f.i], f.survfit$strata[f.i]))
}
#create data.frame with data from survfit (create column for strata)#
f.frame <- data.frame(time=f.survfit$time, n.risk=f.survfit$n.risk, n.event=f.survfit$n.event, n.censor = f.survfit
$n.censor, surv=f.survfit$surv, upper=f.survfit$upper, lower=f.survfit$lower, strata=factor(f.strata))
#remove temporary data#
rm(f.strata)
#create first two rows (start at 1) for each strata#
for(f.i in 1:length(f.survfit$strata)){
#take only subset for this strata from data#
f.subset <- subset(f.frame, strata==names(f.survfit$strata)[f.i])
#create first two rows (time: 0, time of first event)#
f.start <- data.frame(time=c(0, f.subset$time[1]), n.risk=rep(f.survfit[f.i]$n, 2), n.event=c(0,0),
n.censor=c(0,0), surv=c(1,1), upper=c(1,1), lower=c(1,1), strata=rep(names(f.survfit$strata)[f.i],
2))
#add first two rows to dataset#
f.frame <- rbind(f.start, f.frame)
#remove temporary data#
rm(f.start, f.subset)
}
#reorder data#
f.frame <- f.frame[order(f.frame$strata, f.frame$time), ]
#rename row.names#
rownames(f.frame) <- NULL
}
#return frame#
return(f.frame)
}


#define custom function to draw kaplan-meier curve with ggplot#
qplot_survival <- function(f.frame, f.CI="default", f.shape=3){
#use different plotting commands dependig whether or not strata's are given#
if("strata" %in% names(f.frame) == FALSE){
#confidence intervals are drawn if not specified otherwise#
if(f.CI=="default" | f.CI==TRUE ){
#create plot with 4 layers (first 3 layers only events, last layer only censored)#
#hint: censoring data for multiple censoring events at timepoint are overplotted#
#(unlike in plot.survfit in survival package)#
ggplot(data=f.frame) + geom_step(aes(x=time, y=surv), direction="hv") + geom_step(aes(x=time,
y=upper), directions="hv", linetype=2) + geom_step(aes(x=time,y=lower), direction="hv", linetype=2) +
geom_point(data=subset(f.frame, n.censor==1), aes(x=time, y=surv), shape=f.shape)
}
else {
#create plot without confidence intervals#
ggplot(data=f.frame) + geom_step(aes(x=time, y=surv), direction="hv") +
geom_point(data=subset(f.frame, n.censor==1), aes(x=time, y=surv), shape=f.shape)
}
}
else {
if(f.CI=="default" | f.CI==FALSE){
#without CI#
ggplot(data=f.frame, aes(group=strata, colour=strata)) + geom_step(aes(x=time, y=surv),
direction="hv") + geom_point(data=subset(f.frame, n.censor==1), aes(x=time, y=surv), shape=f.shape)
}
else {
#with CI (hint: use alpha for CI)#
ggplot(data=f.frame, aes(colour=strata, group=strata)) + geom_step(aes(x=time, y=surv),
direction="hv") + geom_step(aes(x=time, y=upper), directions="hv", linetype=2, alpha=0.5) +
geom_step(aes(x=time,y=lower), direction="hv", linetype=2, alpha=0.5) +
geom_point(data=subset(f.frame, n.censor==1), aes(x=time, y=surv), shape=f.shape)
}
}
}

绘制全局生存曲线(95% CI):

它没有给出任何错误:

# Kaplan-Meier plot, global survival (with CI)
t.survfit <- survfit(t.Surv~1, data=acbi30)
t.survframe <- createSurvivalFrame(t.survfit)
t.survfit
qplot_survival(t.survframe, TRUE, 20)

绘制分层生存曲线:

给出上面提到的错误:

# Kaplan-Meier plot, stratified survival
t.survfit2 <- survfit(t.Surv~levo, data=acbi30)
t.survframe2 <- createSurvivalFrame(t.survfit2)
t.survfit2
qplot_survival(t.survframe2, TRUE, 20)

不使用 ggplot2 绘制结果:

t.survframe2的结构对我来说似乎没问题,没有任何空行,所以这一定是qplot_survival读取我在t.survframe2中的数据的问题。制作一个简单的绘图不会返回任何错误:

t.survframe2
plot(t.survfit2)

我的数据框问题出在哪里?创建的函数可以很好地与其他数据集配合使用,但不适用于这个数据集...

提前谢谢您,

马雷维夫

session 信息:

sessionInfo()

R版本2.15.2 (2012-10-26) 平台:i386-w64-mingw32/i386(32位)

locale:

[1] LC_COLLATE=Spanish_Spain.1252  LC_CTYPE=Spanish_Spain.1252   
[3] LC_MONETARY=Spanish_Spain.1252 LC_NUMERIC=C                  
[5] LC_TIME=Spanish_Spain.1252    

attached base packages:
[1] grid      splines   stats     graphics  grDevices utils     datasets 
[8] methods   base     

other attached packages:
[1] ggplot2_0.9.3    abind_1.4-0      survival_2.36-14 knitr_0.8       

loaded via a namespace (and not attached):
 [1] colorspace_1.1-1   dichromat_1.2-4    digest_0.5.2      
 [4] evaluate_0.4.2     formatR_0.7        gtable_0.1.2      
 [7] labeling_0.1       MASS_7.3-22        munsell_0.4       
[10] plyr_1.8           proto_0.3-9.2      RColorBrewer_1.0-5
[13] reshape2_1.2.1     scales_0.2.3       stringr_0.6.1     
[16] tools_2.15.2    

最佳答案

我对您的 qplot_survival() 函数做了一些整容手术。主要问题似乎是 geom_pointdata = 参数中的子集条件;在 t.survframet.survframe2 中,n.censor 表格产生值 0、3 和 12。通过更改子集条件到 n.censor > 0,我设法在所有情况下都得到了一个图。我也没有看到 f.CI = "default" 的意义,因此我将默认值设置为 TRUE 并相应地修改了 if 条件。

qplot_survival <- function(f.frame, f.CI= TRUE, f.shape=3)
{
 # use different plotting commands depending whether 
 # or not strata are given#
if(!("strata" %in% names(f.frame)))
{
  #confidence intervals are drawn if not specified otherwise#
   if( isTRUE(f.CI) )
   {
      # create plot with 4 layers (first 3 layers only events, 
      # last layer only censored)#
      # hint: censoring data for multiple censoring events at 
      # timepoint are overplotted#
      # (unlike in plot.survfit in survival package)#
   ggplot(data=f.frame) + 
      geom_step(aes(x=time, y=surv), direction="hv") + 
      geom_step(aes(x=time, y=upper), direction ="hv", linetype=2) + 
      geom_step(aes(x=time,y=lower), direction="hv", linetype=2) +
      geom_point(data=subset(f.frame, n.censor > 0), 
                 aes(x=time, y=surv), shape=f.shape)
   } else {
  #create plot without confidence intervals#
   ggplot(data=f.frame) + 
      geom_step(aes(x=time, y=surv), direction="hv") +
      geom_point(data=subset(f.frame, n.censor > 0), 
                 aes(x=time, y=surv), shape=f.shape)
          }
} else {
  if( !(isTRUE(f.CI)) ){
#without CI#
  ggplot(data=f.frame, aes(group=strata, colour=strata)) + 
     geom_step(aes(x=time, y=surv), direction="hv") + 
     geom_point(data=subset(f.frame, n.censor > 0), 
                aes(x=time, y=surv), shape=f.shape)
} else {

#with CI (hint: use alpha for CI)#
  ggplot(data=f.frame, aes(x = time, colour=strata, group=strata)) + 
      geom_step(aes(y=surv), direction="hv") + 
      geom_step(aes(y=upper), direction="hv", 
                   linetype=2, alpha=0.5) +
      geom_step(aes(y=lower), direction="hv", 
                   linetype=2, alpha=0.5) +
      geom_point(data=subset(f.frame, n.censor > 0), 
                 aes(y=surv), shape=f.shape)
      }
   }
}

进行这些更改后,以下图对我来说都有效:

qplot_survival(t.survframe2, TRUE, 20)
qplot_survival(t.survframe2, FALSE, 20)
qplot_survival(t.survframe, TRUE, 20)
qplot_survival(t.survframe, FALSE, 20)

一些评论:

  1. 函数内部的子集化可能很危险,因为有时(如本例所示)满足条件会返回零行数据帧。我会考虑 geom_point() 层是否真的有必要。
  2. 在一些地方,您在 geom_step() 调用中使用了 directions = "hv"。该参数不是复数的,并且已在上面进行了更改。
  3. 我认为这可以更有效地完成,但是从 survfit 对象中提取感兴趣的列的一种方法是 t.survfit,类似于这个:

(当分层存在时展开比较)

comps <- c(2:6, 8, 10);
t.fit <- as.data.frame(do.call(cbind, lapply(comps, function(j) t.survfit[[j]])))
names(t.fit) <- names(t.survfit)[comps]

关于r - ggplot2 绘制 Kaplan-Meier 图时在生存数据框中找到空行,我们在Stack Overflow上找到一个类似的问题: https://stackoverflow.com/questions/14160928/

相关文章:

节点搜索的图遍历

graphics - 打印和图形[]

python - 图 - Python - 存储由图组成的图数据库 - 图数据库理论

使用 lubridate 检索会计年度中的月份数

r - 在矩阵中删除NA/NaN/Inf

c# - DrawLine 方法提供比 DrawLines 方法更高的质量

java - 如何在 Java3D 中旋转一组对象

C崩溃,大图上的BFS可能路线

r - 绘制修改后的点和线图 - r 中的可变性为 "spike"图

r - 如何通过现有名称列表将矩阵行或列名称更改为新名称?