Neyman's confidence intervals make no attempt to provide coverage of the parameter in the case of any particular interval. Instead they provide coverage over all possible parameter values in the long run. In a sense they attempt to be globally accurate at the expense of local accuracy.
Confidence intervals for binomial proportions offer a clear illustration of this issue. Neymanian assessment of intervals yields the irregular coverage plots like this, which is for 95% Clopper-Pearson intervals for n=10 Binomial trials:
![Clopper-Pearson coverage plot](https://i.stack.imgur.com/Xg9e0.png)
There is an alternative way to do coverage, one that I personally think is much more intuitively approachable and (thus) useful. The coverage by intervals can be specified conditional on the observed result. That coverage would be local coverage. Here is a plot showing local coverage for three different methods of calculation of confindence intervals for binomial proportions: Clopper-Pearson, Wilson's scores, and a conditional exact method that yield intervals identical to Bayesian intervals with a uniform prior:
![Conditional coverage for three types of interval](https://i.stack.imgur.com/ryXNu.png)
Notice that the 95% Clopper-Pearson method gives over 98% local coverage but the exact conditional intervals are, well, exact.
A way to think of the difference between the global and local intervals is to consider the global to be inversions of Neyman-Pearson hypothesis tests where the outcome is a decision that is made on the basis of consideration of long-term error rates for the current experiment as a member of the global set of all experiments that might be run. The local intervals are more akin to inversion of Fisherian significance tests which yield a P value which represents evidence against the null in from this particular experiment.
(As far as I know, the distinction between global and local statistics was first made in an unpublished Master’s thesis by Claire F Leslie (1998) Lack of confidence : a study of the suppression of certain counter-examples to the Neyman-Pearson theory of statistical inference with particular reference to the theory of confidence intervals. That thesis is held by the Baillieu library at The University of Melbourne.)
If the bootstrapping procedure and the formation of the confidence interval were performed correctly, it means the same as any other confidence interval. From a frequentist perspective, a 95% CI implies that if the entire study were repeated identically ad infinitum, 95% of such confidence intervals formed in this manner will include the true value. Of course, in your study, or in any given individual study, the confidence interval either will include the true value or not, but you won't know which. To understand these ideas further, it may help you to read my answer here: Why does a 95% Confidence Interval (CI) not imply a 95% chance of containing the mean?
Regarding your further questions, the 'true value' refers to the actual parameter of the relevant population. (Samples don't have parameters, they have statistics; e.g., the sample mean, $\bar x$, is a sample statistic, but the population mean, $\mu$, is a population parameter.) As to how we know this, in practice we don't. You are correct that we are relying on some assumptions--we always are. If those assumptions are correct, it can be proven that the properties hold. This was the point of Efron's work back in the late 1970's and early 1980's, but the math is difficult for most people to follow. For a somewhat mathematical explanation of the bootstrap, see @StasK's answer here: Explaining to laypeople why bootstrapping works . For a quick demonstration short of the math, consider the following simulation using R
:
# a function to perform bootstrapping
boot.mean.sampling.distribution = function(raw.data, B=1000){
# this function will take 1,000 (by default) bootsamples calculate the mean of
# each one, store it, & return the bootstrapped sampling distribution of the mean
boot.dist = vector(length=B) # this will store the means
N = length(raw.data) # this is the N from your data
for(i in 1:B){
boot.sample = sample(x=raw.data, size=N, replace=TRUE)
boot.dist[i] = mean(boot.sample)
}
boot.dist = sort(boot.dist)
return(boot.dist)
}
# simulate bootstrapped CI from a population w/ true mean = 0 on each pass through
# the loop, we will get a sample of data from the population, get the bootstrapped
# sampling distribution of the mean, & see if the population mean is included in the
# 95% confidence interval implied by that sampling distribution
set.seed(00) # this makes the simulation reproducible
includes = vector(length=1000) # this will store our results
for(i in 1:1000){
sim.data = rnorm(100, mean=0, sd=1)
boot.dist = boot.mean.sampling.distribution(raw.data=sim.data)
includes[i] = boot.dist[25]<0 & 0<boot.dist[976]
}
mean(includes) # this tells us the % of CIs that included the true mean
[1] 0.952
Best Answer
The 95% is not numerically attached at all to how confident you are that you've covered the true effect in your experiment. Perhaps recognizing that "interval using 95% coverage range calculation" might be a more accurate name for it. You can make the choice to decide that the interval contains the true value; and you'll be right if you do that consistently 95% of the time. But you really don't know how likely it is for your particular experiment without more information.
Q1: Your first query conflates two things and misuses a term. No wonder you're confused. A narrower confidence interval may be more precise but, when calculated the same way, such as the 95% method, they all have the same accuracy. They capture the true value the same proportion of the time.
Also, just because it's narrow doesn't mean you're less likely to encounter a sample that falls within that narrow confidence interval. A narrow confidence interval can be achieved one of three ways. The experimental method or nature of the data could just have very low variance. The confidence interval around the boiling point of tap water at sea level is pretty small, regardless of the sample size. The confidence interval around the average weight of people might be rather large because people are very variable but one can make that confidence interval smaller by just acquiring more observations. In that case, as you gain more certainty about where you believe the true value is, by collecting more samples and making a narrower confidence interval, then the probability of encountering an individual in that confidence interval does go down. (it goes down in any case when you increase sample size, but you may not bother collecting the big sample in the boiling water case). Finally, it could be narrow because your sample is unrepresentative. In that case you are actually more likely to have one of the 5% of intervals that does not contain the true value. It's a bit of a paradox regarding CI width and something you should check by knowing the literature and how variable this data typically is.
Further consider that the confidence interval is about trying to estimate the true mean value of the population. If you knew that spot on then you'd be even more precise (and accurate) and not even have a range of estimates. But your probability of encountering an observation with that exact same value would be far lower than finding one within any particular sample based CI.
Q2: A 99% confidence interval is wider than a 95%. Therefore, it's more likely that it will contain the true value. See the distinction above between precise and accurate, you're conflating the two. If I make a confidence interval narrower with lower variability and higher sample size it becomes more precise, the likely values cover a smaller range. If I increase the coverage by using a 99% calculation it becomes more accurate, the true value is more likely to be within the range.