使用ggplot2在直方图上绘制置信区间

Question

The code below plots of the sampling distribution of the mean and calculate 20 lots of 95% confidence intervals. How can I plot the confidence intervals on the histogram, as in the Photoshopped image below?

# plot sampling distribution of mean -----------------------------------------------------------
set.seed(1)

population <- rnorm(10000, 3, 3)

population_mean <- mean(population)

my_sample <- sample(population, 100, replace = FALSE)

standard_error <- sqrt(var(my_sample)/length(my_sample))

sampling_distribution_of_mean <- rnorm(10000, mean = population_mean, sd = standard_error)

library(ggplot2)
ggplot(data.frame(x = sampling_distribution_of_mean), aes(x)) + geom_histogram() + geom_vline(xintercept = population_mean, color = "red")


# calculate 20 lots of 95% confidence intervals -----------------------------------------------------------

my_confidence_intervals <- function(){

    my_sample <- sample(population, 100, replace = FALSE)

    sample_mean <- mean(my_sample)

    standard_error <- sqrt(var(my_sample)/length(my_sample))

    margin_of_error <- 1.96*standard_error

    mean_minus_margin_of_error <- sample_mean - margin_of_error
    mean_plus_margin_of_error <- sample_mean + margin_of_error

    c(mean_minus_margin_of_error, mean_plus_margin_of_error)

}

library(plyr)
llply(1:20, function(x) my_confidence_intervals())

Answer 1

You would want to build a data.frame containing the intervals and then add a layer of horizontal error bars to plot them. First, i transform your ranges into a data.frame

xx<-llply(1:20, function(x) my_confidence_intervals())
xx<-data.frame(y=1:20*50, x=do.call(rbind, xx))

Now I add them to the plot

ggplot(data.frame(x = sampling_distribution_of_mean), aes(x)) + 
    geom_histogram() + 
    geom_vline(xintercept = population_mean, color = "red") + 
    geom_errorbarh(aes(y=y, x=x.1, xmin=x.1, xmax=x.2), data=xx, col="#0094EA", size=1.2)

which gives

Notice that i explicitly set y-values for each of the ranges when creating the data.frame.