Computing Equal-Tailed Credible Intervals
You'll learn how to calculate a credible interval by symmetrically trimming equal probabilities from both tails of your posterior distribution.
The Core Idea
An equal-tailed credible interval places exactly the same probability mass in each tail outside the interval. For a 95% credible interval, you exclude 2.5% from the lower tail and 2.5% from the upper tail, capturing the middle 95% of the posterior distribution.
Think of it like trimming a hedge: you cut the same amount from both sides to keep it balanced and centered.
How to Compute It
Once you have your posterior distribution, computing an equal-tailed credible interval is straightforward:
- Decide your credibility level (e.g., 95%)
- Calculate the tail probabilities: For 95%, that's (1 - 0.95) / 2 = 0.025 in each tail
- Find the quantiles: The lower bound is the 2.5th percentile, the upper bound is the 97.5th percentile
For a Beta posterior distribution (from the Beta-Binomial model), you would find the values where the cumulative distribution function equals 0.025 and 0.975.
Why "Equal-Tailed"?
The symmetry makes these intervals intuitive and easy to communicate. They answer: "Where does the middle X% of my posterior probability live?" This is the most common approach in practice, though alternatives like highest posterior density intervals exist for asymmetric distributions.
Key Takeaway: An equal-tailed credible interval is computed by finding the quantiles of the posterior distribution that leave equal probability in both tails—for a 95% interval, use the 2.5th and 97.5th percentiles.