Chapter 50: Evaluating Confidence Intervals: Length and Coverage Probability

There are two ways to assess confidence intervals. The first is the coverage probability, or the probability that the confidence interval captures the parameter. For exact confidence intervals, the coverage probability is equal to the target coverage probability (if the model is correct). For example, the confidence interval for the mean of a normal population, created using a pivotal quantity with a t-density, is exact; therefore, if the population is truly normal, a 95% confidence interval will capture the population mean exactly 95% of the time.