Homepage > MCMC
Markov Chain Monte Carlo
  • Concept of Distribution
  • Monte Carlo Simulation
  • Markov Chain
  • Metropolis Algorithm
    Gibbs Sampler
  • Dynamics of Gibbs Models
  • Gibbs Distribution
  • Ising and Potts Models
    Bayesian Models
  • Joint Distribution
  • Frequentist Approach
  • Bayesian Approach
  • Dirichlet Distribution
  • Mixture Model
  • MCMC scheme
    Data Mining and MCMC
  • Relative Report Rate
  • Graphical Model
  • MCMC scheme
    Explore MCMC with R
  • Normal Mixture
  • Searching for Maxima
  • Bernoulli Trials
  • Random Beta Walk
  • Metropolis Algorithm
  • Potts Model

    • Bayesian Models

    Bayesian uses the concept of prior belief about the parameter $ \theta$ of interest. Then the uncertainty of $ \theta$ changes according to the data $ \mathcal{D}$ . Here Bayesian interprets $ \theta$ as a random variable, and the prior belief is given in the form of probability density $ \pi(\theta)$ of $ \theta$ . In a Bayesian model we will investigates the postrior density $ \pi(\theta\:\vert\:\mathcal{D})$ of $ \theta$ .

    Conditional probability. Let $ A$ and $ B$ be two events where $ P(B) \neq 0$ . Then the conditional probability of $ A$ given $ B$ can be defined as

    $\displaystyle P(A\vert B) = \frac{P(A\cap B)}{P(B)}. $

    The idea of ``conditioning'' is that ``if we have known that $ B$ has occurred, the sample space should have become $ B$ .'' It is often the case that one can use $ P(A\vert B)$ to find

    $\displaystyle P(A\cap B) = P(A\vert B)P(B). $

    Law of total probability. Let $ A$ and $ B$ be two events. Then we can write the probability $ P(A)$ as

    $\displaystyle P(A) = P(A \cap B) + P(A \cap B^c) = P(A \vert B) P(B) + P(A \vert B^c) P(B^c). $

    In general, suppose that we have a sequence $ B_1, B_2, \ldots, B_n$ of mutually disjoint events satisfying $ \bigcup_{i=1}^n B_i = \Omega$ , where ``mutual disjointness'' means that $ B_i \cap B_j = \emptyset$ for all $ i \neq j$ . (The events $ B_1, B_2, \ldots, B_n$ are called ``a partition of $ \Omega$ .'') Then for any event $ A$ we have

    $\displaystyle P(A) = \sum_{i = 1}^n P(A \cap B_i) = \sum_{i = 1}^n P(A\vert B_i)P(B_i). $

    Bayes rule. Let $ A$ and $ B_1, \ldots, B_n$ be events such that the $ B_i$ 's are mutually disjoint, $ \bigcup_{i=1}^n B_i = \Omega$ and $ P(B_i) > 0$ for all $ i$ . Then

    $\displaystyle P(B_j\vert A) = \frac{P(A \cap B_j)}{P(A)} = \frac{P(A\vert B_j)P(B_j)}{\sum_{i=1}^n P(A\vert B_i)P(B_i)} $

    is called Bayes rule.

    Concept of Independence. Intuitively we would like to say that $ A$ and $ B$ are independent if knowing about one event give us no information about another. That is, $ P(A\vert B) = P(A)$ and $ P(B\vert A) = P(B)$ . We say $ A$ and $ B$ are independent if

    $\displaystyle P(A\cap B) = P(A)P(B). $

    This definition is symmetric in $ A$ and $ B$ , and allows $ P(A)$ and/or $ P(B)$ to be 0. Furthermore, a collection of events $ A_1, A_2, \ldots, A_n$ is said to be mutually independent if it satisfies

    $\displaystyle P(A_{i_1}\cap\cdots\cap A_{i_m}) = P(A_{i_1}) P(A_{i_2}) \cdots P(A_{i_m}) $

    for any subcollection $ A_{i_1}, A_{i_2}, \ldots, A_{i_m}$ .

    © TTU Mathematics