返回题库

硬币偏置的贝叶斯估计

Bayesian Estimate of a Coin's Bias

专题
Probability / 概率
难度
L4

题目详情

一枚硬币正面概率为未知参数 θ\theta。你对 θ\theta 的先验为 [0,1][0,1] 上的均匀分布。

掷硬币 10 次,观察到 6 次正面。求 θ\theta 的贝叶斯估计(后验均值)。

You have a coin with an unknown probability of heads, θ\theta. You assume a uniform prior distribution for θ\theta over [0,1][0,1]. After tossing the coin 10 times, you observe 6 heads. What is the Bayesian estimate (posterior mean) of θ\theta?

解析

均匀先验等价于 θBeta(1,1)\theta\sim\mathrm{Beta}(1,1)

观察到 6 次正面、4 次反面后,后验为

θdataBeta(1+6, 1+4)=Beta(7,5).\theta\mid \text{data}\sim \mathrm{Beta}(1+6,\ 1+4)=\mathrm{Beta}(7,5).

Beta(α,β)(\alpha,\beta) 的均值为 α/(α+β)\alpha/(\alpha+\beta),因此

E[θdata]=77+5=7120.5833.\mathbb{E}[\theta\mid \text{data}]=\frac{7}{7+5}=\boxed{\frac{7}{12}}\approx 0.5833.