Markov chains: finite sink is recurrent (incomplete)
Dependencies: (incomplete)
- Markov chain
- Markov chains: recurrent states
- Markov chains: accessibility and state classes
- Markov chains: recurrence is a class property
- Markov chains: recurrent iff expected number of visits is infinite
$\newcommand{\E}{\operatorname{E}}$ Let $X = [X_0, X_1, \ldots]$ be a markov chain having state space $D$. Let $I$ be a state class containing a finite number of states. Then $I$ is recurrent if no other state class is accessible from $I$.
Proof
For any state $i$, let $N_i$ be the number of times the state becomes $i$. Formally, let $N_i = \sum_{t=0}^{\infty} \begin{cases}1&\textrm{ if }X_t=i \\0&\textrm{ otherwise}\end{cases}$.
Let $I = \{i_1, i_2, \ldots, i_r\}$. Let $j \not\in I$. Then $j$ is not accessible from $i_1$, so $\Pr(N_j = 0 \mid X_0 = i_1) = 1$. Hence, $\E(N_j \mid X_0 = i_1) = 0$. However, $\sum_{i \in D} N_i = \infty$, so $\E(N_{i_k} \mid X_0 = i_1) = \infty$ for some $k$. (incomplete proof)
Dependency for: None
Info:
- Depth: 10
- Number of transitive dependencies: 32
Transitive dependencies:
- /analysis/topological-space
- /sets-and-relations/countable-set
- /sets-and-relations/poset
- /sets-and-relations/equivalence-relation
- /sets-and-relations/de-morgan-laws
- /measure-theory/linearity-of-lebesgue-integral
- /measure-theory/lebesgue-integral
- σ-algebra
- Generated σ-algebra
- Borel algebra
- Measurable function
- Generators of the real Borel algebra (incomplete)
- Measure
- σ-algebra is closed under countable intersections
- Group
- Ring
- Field
- Vector Space
- Semiring
- Matrix
- Identity matrix
- Probability
- Conditional probability (incomplete)
- Random variable
- Markov chain
- Markov chains: recurrent states
- Chapman-Kolmogorov equation
- Markov chains: accessibility and state classes
- Expected value of a random variable
- Linearity of expectation
- Markov chains: recurrent iff expected number of visits is infinite
- Markov chains: recurrence is a class property