在geom_histogram或stat_bin上覆蓋geom_points

Question

我想使用ggplot繪製直方圖（或使用stat_bin的步驟圖），並使用geom_point覆蓋幾個點。

這裏是一個base實現：

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) 

這給：

然而，當我嘗試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") 

的點和直方圖是錯位：

理想我想用一步情節來繪製它：

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") 

它看起來非常像geom_histogram

而且我也想有線的末端觸摸y = 0線。

所以我可以在stat_bin的步驟圖中正確顯示嗎？

Answer 1

你selected.df的y.values與breakshist()用途方面，但geom_histogram()使用另一個breaks。 （geom_histogram(bins)不等同於hist(breaks)）。此外，在步驟圖中，上下發生在其breaks的中間值上。 ggplot_build(gg.obj)$data（或plot(gg.obj)$data）爲您提供一些信息，中斷，計數等。

geom_histgram
的方法是基本相同base.plot。如果您需要與base.plot相同的輸出，請使用breaks = h$breaks而不是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
您可以用同樣的方法繪製成geom_histgram。重要的一點是上下並不是休息而是中間值。

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")