Bayesian information criterion

Pocket

If sample size (n) was so large enough, Bayesian information criterion (BIC), an evaluation criteria of model estimated by maximum likelihood method, would be approximated with Laplace method by integrating marginal likelihood corresponding to posterior probability of model. θ is a parameter of p-dimension and f(xn|θ) is probability distribution function, respectively.

\displaystyle BIC = -2\log f(x_n|\hat\theta) + p\log n

References:
Probability density function, expected value and variance of each probability distribution
How to calculate Akaike information criterion with probability distribution function?

Pocket

投稿者: admin

趣味:写真撮影とデータベース. カメラ:TOYO FIELD, Hasselblad 500C/M, Leica M6. SQL Server 2008 R2, MySQL, Microsoft Access.

コメントを残す

メールアドレスが公開されることはありません。 * が付いている欄は必須項目です