Confidence Interval (CI) simulation in R: How?

问题:

I was wondering how I could check via simulation in R that the 95% Confidence Interval obtained from a binomial test with 5 successes in 15 trials when TRUE p = .5 has a 95% "Coverage Probability" in the long-run?

Here is the 95% CI for such a test using R (how can show that the following CI has a 95% coverage in the long-run if TRUE p = .5):

as.numeric(binom.test(x = 5, n = 15, p = .5)[[4]])
# > [1] 0.1182411 0.6161963 (in the long-run 95% of the time, ".5" is contained within these
#                            two numbers, how to show this in R?)

回答1:

Something like this?

fun <- function(n = 15, p = 0.5){
    x <- rbinom(1, size = n, prob = p)
    res <- binom.test(x, n, p)[[4]]
    c(Lower = res[1], Upper = res[2])
}

set.seed(3183)
R <- 10000
sim <- t(replicate(R, fun()))

Note that binom.test when called with 5 successes, 15 trials and p = 0.5 will always return the same value, hence the call to rbinom. The number of successes will vary. We can compute the proportion of cases when p is between Lower and Upper.

cov <- mean(sim[,1] <= .5 & .5 <= sim[,2])