在此直方图中将计数更改为概率

Question

我正在努力将计数更改为以下直方图中的概率，而不会弄乱红色区域。另外，如何将 1,10 与 x 轴上的其余数字对齐？

library(dplyr)
library(tibble)
library(ggplot2)
nrep = 10000
scientific <- function(x){
  ifelse(x==1e+0, "1", ifelse(x==1e+1,"10",parse(text=gsub("[+]", "", gsub("1e+", "10^", scales::scientific_format()(x))))))
}
bw <- 0.05
mx=rf(nrep,5,2)
df = tibble(x = mx)
ggplot(df,aes(x)) + 
geom_histogram(binwidth=bw, color="white", fill = "#1380A1") + 
geom_histogram(data=df %>% filter(x < 10^(-1) + 1.15*bw), binwidth=bw, color="white", fill = "red") +
geom_density(aes(y = bw*after_stat(count)), color="blue") +
scale_x_continuous(trans="log10", breaks = 10^seq(-1, 5, by = 1), labels = scientific)

Answer 1

您需要将层geom_histogram 中的比例从计数更改为密度。在第一个直方图中，您可以使用after_stat(density) 来实现，它相当于after_stat(count/sum(count))/bw。但是，相同的过程在第二个直方图中不起作用，因为当您对数据集进行子集化时，sum(count) 是不同的。如果这样做，第二个直方图将采用不同的比例。

library(dplyr)
library(tibble)
library(ggplot2)
nrep = 10000
scientific <- function(x){
  ifelse(x==1e+0, "1", ifelse(x==1e+1,"10",parse(text=gsub("[+]", "", gsub("1e+", "10^", scales::scientific_format()(x))))))
}
bw <- 0.05
mx=rf(nrep,5,2)

df = tibble(x = mx) 
pdf <- df %>% filter(x < 10^(-1) + 1.15*bw)
ggplot() + 
  geom_histogram(data = df,
                 aes(x = x, y = after_stat(count/sum(count)/bw),
                 binwidth=bw, color="white", fill = "#1380A1") + 
  geom_histogram(data = pdf, 
                 aes(x = x, y = after_stat(count/sum(count)/bw),
                 binwidth=bw, color="white", fill = "red") +
  geom_density(data = df, 
               aes(x = x), color="blue") +
  scale_x_continuous(trans="log10", breaks = 10^seq(-1, 5, by = 1), 
                     labels = scientific)

因此，需要从第一个直方图计算出同分母的密度，定义为nrep。

library(dplyr)
library(tibble)
library(ggplot2)
nrep = 10000
scientific <- function(x){
  ifelse(x==1e+0, "1", ifelse(x==1e+1,"10",parse(text=gsub("[+]", "", gsub("1e+", "10^", scales::scientific_format()(x))))))
}
bw <- 0.05
mx=rf(nrep,5,2)

df = tibble(x = mx) 
pdf <- df %>% filter(x < 10^(-1) + 1.15*bw)
ggplot() + 
  geom_histogram(data = df,
                 aes(x = x, y = after_stat(density)),
                 binwidth=bw, color="white", fill = "#1380A1") + 
  geom_histogram(data = pdf, 
                 aes(x = x, y = after_stat(count/nrep)/bw),
                 binwidth=bw, color="white", fill = "red") +
  geom_density(data = df, 
               aes(x = x), color="blue") +
  scale_x_continuous(trans="log10", breaks = 10^seq(-1, 5, by = 1), 
                     labels = scientific)