Course contentsShow
← Mathematics
Lesson 3115 of 3,12160. Statistics — ModelingFree lesson

Conjugate Priors: Normal-Normal Model

Explore Bayesian inference for a normal mean with known variance using a normal prior to obtain a normal posterior.

Conjugate Priors: Normal-Normal Model

You'll learn how to use a normal prior distribution for a normal mean (with known variance) to produce a normal posterior—a mathematically elegant Bayesian update.

The Core Idea

In Bayesian inference, we often want to estimate the mean μ of a normal distribution when we already know the variance σ². If we choose a normal prior for μ, the posterior distribution (after observing data) is also normal. This "prior-posterior conjugacy" makes calculations clean and interpretable.

Think of it like this: you have a prior belief about where the true mean lies (your prior normal distribution). After seeing data, you combine that belief with the evidence, and your updated belief (posterior) keeps the same normal shape—just with updated parameters.

The Model

  • Likelihood: Your data X₁, X₂, ..., Xₙ come from Normal(μ, σ²) with σ² known
  • Prior: You believe μ ~ Normal(μ₀, τ₀²) before seeing data
  • Posterior: After observing data, μ | data ~ Normal(μₙ, τₙ²)

The posterior parameters are:

τₙ² = 1 / (1/τ₀² + n/σ²)          [posterior variance]
μₙ = τₙ² × (μ₀/τ₀² + nX̄/σ²)       [posterior mean]

Where X̄ is the sample mean. Notice the posterior mean is a weighted average of your prior mean μ₀ and the sample mean X̄, with weights determined by their relative precisions (1/variance).

Why This Matters

This conjugate relationship lets you update beliefs sequentially as data arrives, always working with tractable normal distributions. The math stays closed-form—no numerical integration needed.

Key Takeaway: When you have normal data with known variance and use a normal prior for the mean, your posterior is also normal, with parameters blending prior knowledge and observed evidence according to their relative precision.