这 贝叶斯 因子是两个相互竞争的假设(通常是零假设 (M₁) 和备择假设 (M₂))的边际似然比。它量化了在给定观测数据 D 的情况下,一个假设相对于另一个假设的支持程度。公式为 K = P(D|M₁) / P(D|M₂)。K > 1 表示数据更倾向于 M₁ 而非 M₂。

(图片仅供参考)
这 贝叶斯 因子是两个相互竞争的假设(通常是零假设 (M₁) 和备择假设 (M₂))的边际似然比。它量化了在给定观测数据 D 的情况下,一个假设相对于另一个假设的支持程度。公式为 K = P(D|M₁) / P(D|M₂)。K > 1 表示数据更倾向于 M₁ 而非 M₂。
这 贝叶斯 factor is the Bayesian alternative to the frequentist p-value for hypothesis testing. Unlike a p-value, which only provides evidence against the null hypothesis, the Bayes factor can quantify evidence for the null, for the alternative, or indicate that the data is uninformative. The magnitude of the Bayes factor provides a continuous scale of evidence. For instance, a Bayes factor of 10 is often considered ‘strong’ evidence for one model over another, while a value between 1 and 3 is considered ‘anecdotal’ or ‘weak’ evidence.
The core component of the Bayes factor is the marginal likelihood, [latex]P(D|M) = \int P(D|\theta, M)P(\theta|M) d\theta[/latex]. This is the probability of the observed data averaged over the prior distribution of the parameters [latex]\theta[/latex] for a given model [latex]M[/latex]. This integral makes the Bayes factor sensitive to the choice of prior distributions, which is a point of contention and active research. It also makes it computationally challenging to calculate, often requiring numerical methods like MCMC 或者使用贝叶斯信息准则(BIC)等近似方法。尽管存在这些挑战,但它能够权衡相互竞争的假设的证据,使其成为科学推断和模型选择的强大工具。
贝叶斯因子
(如果日期未知或不相关,例如“流体力学”,则提供其显著出现的近似估计)
只有注册会员才能免费获取完整尺寸的图片和下载内容。
> 登录 <