覆盖geom_histogram或stat_bin上的geom_points

Question

我想使用ggplot绘制直方图(或使用的步骤图stat_bin)并使用它覆盖几个点geom_point.

这是一个base实现:

library(plotrix)
set.seed(10)
df <- data.frame(id=LETTERS,val=rnorm(length(LETTERS)))
selected.ids <- sample(LETTERS,3,replace=F)
h <- hist(df$val,plot=F,breaks=10)
cols <- sapply(rainbow(length(selected.ids)),function(x) color.id(x)[1])
selected.df <- data.frame(id=selected.ids,col=cols,stringsAsFactors=F)
selected.df$x <- df$val[which(df$id %in% selected.ids)]
selected.df <- selected.df[order(selected.df$x),]
selected.df$y <- h$counts[findInterval(selected.df$x,h$breaks)]
selected.df$col <- factor(selected.df$col,levels=cols)
plot(h)
segments(x0=selected.df$x,x1=selected.df$x,y0=selected.df$y,y1=selected.df$y,cex=18,lwd=8,col=selected.df$col)

这使:

但是,当我尝试ggplot:

ggplot(df,aes(x=val))+geom_histogram(bins=10,colour="black",alpha=0,fill="#FF6666")+geom_point(data=selected.df,aes(x=x,y=y,colour=factor(col)),size=2)+scale_fill_manual(values=levels(selected.df$col),labels=selected.df$id,name="id")+scale_colour_manual(values=levels(selected.df$col),labels=selected.df$id,name="id")

点和直方图未对齐:

理想情况下,我想使用步骤图绘制它:

ggplot(df,aes(x=val))+stat_bin(geom="step",bins=10)+geom_point(data=selected.df,aes(x=x,y=y,colour=factor(col)),size=2)+scale_fill_manual(values=levels(selected.df$col),labels=selected.df$id,name="id")+scale_colour_manual(values=levels(selected.df$col),labels=selected.df$id,name="id")

这看起来非常像 geom_histogram

而且我也想让线的末端触及y = 0线.

所以我使用stat_bin在步骤图中得到正确的结果？

Answer 1

您selected.df的y.values是由使用制成的breaks hist()，但geom_histogram()使用了另一个breaks。（geom_histogram(bins)不等于hist(breaks)只是确定）。此外，在阶梯图中，上下发生在其 的中间值上breaks。ggplot_build(gg.obj)$data(或plot(gg.obj)$data) 为您提供一些信息、中断、计数等。

geom_histgram 的
方式与 基本相同base.plot。如果您想要与 相同的输出base.plot，请使用breaks = h$breaks代替bars = 10。

# a common part to base and ggplot2
library(plotrix)
set.seed(10)
df <- data.frame(id = LETTERS, val = rnorm(length(LETTERS)))
selected.ids <- sample(LETTERS, 3, replace = F)
cols <- sapply(rainbow(length(selected.ids)), function(x) color.id(x)[1])
selected.df <- data.frame(id=selected.ids, col=cols, stringsAsFactors = F)
selected.df$x <- df$val[which(df$id %in% selected.ids)]
selected.df <- selected.df[order(selected.df$x),]
selected.df$col <- factor(selected.df$col, levels=cols)

# (1) make a histogram
g <- ggplot(df, aes(x = val)) + geom_histogram(bins = 10, colour = "black", alpha = 0, fill = "#FF6666")
  # base; h <- hist(df$val, plot = F, breaks = 10)

# (2) get its breaks
g.data <- ggplot_build(g)$data[[1]]
g.breaks <- c(g.data$xmin, tail(g.data$xmax, n=1))
  # base; h$breaks

# (3) get counts of specific x values
selected.df$y <- g.data$count[findInterval(selected.df$x, g.breaks)]
  # base; selected.df$y <- h$counts[findInterval(selected.df$x,h$breaks)]

# (4) draw
g + geom_point(data = selected.df, aes(x = x, y = y, colour = factor(col)), size = 2) + 
  scale_fill_manual(values = levels(selected.df$col), labels = selected.df$id,name = "id") + 
  scale_colour_manual(values = levels(selected.df$col), labels = selected.df$id, name = "id")

stat_bin
可以用和 一样的方法来绘制geom_histgram。重要的一点是上下不是发生在中断处而是发生在中间值处。

selected.df2 <- selected.df

# (1) make a step plot
s <- ggplot(df, aes(x = val)) + stat_bin(geom = "step", bins = 10)

# (2) get breaks and its middle values
s.data <- ggplot_build(s)$data[[1]]
s.breaks <- c(s.data$xmin, tail(s.data$xmax, n=1))
s.mid.breaks <- s.data$x

# (3) get counts of specific x values using middle values of breaks.
selected.df2$y <- s.data$count[findInterval(selected.df2$x, s.mid.breaks)]

# (4) add a new levels into breaks to start and end at y=0
s.add.breaks <- c(s.breaks[1] - 1.0E-6,    # making lower levels is easy
                  s.breaks, 
                  tail(s.breaks, n=1) + diff(s.breaks[1:2])) # upper need the same range

# (5) draw
ggplot(df, aes(x = val)) + stat_bin(geom = "step", breaks = s.add.breaks) +
  geom_point(data = selected.df2, aes(x = x, y = y, colour = factor(col)), size = 2) +
  scale_fill_manual(values = levels(selected.df2$col), labels = selected.df2$id, name = "id") + 
  scale_colour_manual(values = levels(selected.df2$col), labels = selected.df2$id, name="id")