練習問題を解くことで
著者が何を読者に伝えたかったのかがようやく理解できた気がする。

おまけ：
library(binom)
method2credibility <- function(
Method = c("agresti-coull", "asymptotic","bayes","cloglog","exact","logit",
"probit", "profile","lrt","prop.test","wilson")){
n_values = c(10, 25, 50,100,1000)
p_values = c(0.05, 0.25, 0.5,0.75,0.95)
np2coverage <- function(n,p,m=1000,method=Method){
y=rbinom(m,n,p)
re=binom::binom.confint(y,n,method = Method)
lo=re["lower"]
up=re["upper"]
mean(lo < p & p < up)
}
np2coverage=Vectorize(np2coverage)
re=outer(n_values,p_values,np2coverage)
rownames(re)=n_values
colnames(re)=p_values
re
}

method2credibility("asymptotic") # 正規分布近似
method2credibility("bayes") # ベイズ:事前分布β(0.5,0.5)
method2credibility("exact") # Clopper-Pearson