Normal-Normal Conjugacy
You'll learn how the Normal distribution acts as its own conjugate prior when estimating a mean with known variance, making posterior calculations analytically simple.
The Core Idea
When you're trying to estimate the mean of a normally distributed population (with known variance), and you express your prior belief about that mean as another Normal distribution, Bayes' theorem produces a posterior that is also Normal. This is Normal-Normal conjugacy: Normal prior + Normal likelihood = Normal posterior.
This elegant relationship gives you closed-form formulas—no numerical integration required.
How It Works
Suppose you believe the true mean μ has a Normal prior with mean μ₀ and variance σ₀². You then observe data that follows a Normal likelihood with known variance σ². The posterior distribution for μ is Normal with:
- Posterior mean: a weighted average of your prior mean and the sample mean, weighted by their precisions (inverse variances)
- Posterior variance: combines information from both the prior and the data
The more data you collect (higher precision from likelihood), the more the posterior mean shifts toward the sample mean. The more confident your prior (higher prior precision), the more it anchors the posterior.
Why It Matters
This setup is common when estimating means of measurements with known measurement error—think quality control, sensor readings, or test scores. The conjugacy means you can update beliefs instantly as new observations arrive, without expensive computations.
Like Beta-Binomial for proportions and Gamma-Poisson for rates, Normal-Normal conjugacy lets you focus on interpretation rather than computation.
Key Takeaway: Normal-Normal conjugacy provides closed-form Bayesian updates for mean estimation with known variance, blending prior belief and data according to their relative precision.