Computing the Posterior Distribution
You'll learn: How to actually calculate a posterior beta distribution by mathematically combining your prior belief with observed binomial data.
The Computation Formula
Now that you understand the beta-binomial conjugate pair, it's time to compute the posterior yourself. The beauty of conjugacy is that the math stays simple and closed-form.
When you have:
- A Beta(α, β) prior representing your initial belief
- Binomial data: k successes in n trials
Your posterior is simply:
Posterior = Beta(α + k, β + n - k)
That's it! You add the number of successes to α, and add the number of failures to β.
Why This Works
Think of α as "prior successes" and β as "prior failures." When you observe k real successes and (n - k) real failures, you're literally adding evidence to your prior counts. The beta distribution gracefully absorbs this new information and updates its shape.
Example Calculation
Suppose you have a Beta(2, 8) prior (you think a conversion rate is probably low). You run an experiment and observe 15 successes in 50 trials.
Your posterior is:
- α_new = 2 + 15 = 17
- β_new = 8 + (50 - 15) = 8 + 35 = 43
So your posterior is Beta(17, 43). Your belief has shifted upward because the data showed more success than your prior suggested.
The Intuition
The prior provides "virtual data." Strong priors (large α and β) require more real data to shift beliefs. Weak priors (small α and β) let data dominate quickly. This posterior becomes your new, updated belief about the conversion rate.
Key Takeaway: Computing a posterior beta distribution is simple addition—add observed successes to α and observed failures to β, creating Beta(α + k, β + n - k).