Markov chain, class of positive states of a

A set $ K $ of states of a homogeneous Markov chain $ \xi ( t) $ with state space $ S $ such that the transition probabilities

$$ p _ {ij} ( t) = {\mathsf P} \{ \xi ( t) = j \mid \xi ( 0) = i \} $$

of $ \xi ( t) $ satisfy

$$ \sup _ { t } p _ {ij} ( t) > 0 \ \ \textrm{ for any } i , j \in K , $$

$ p _ {il} ( t) = 0 $ for any $ i \in K $, $ l \in S \setminus K $, $ t > 0 $, and

$$ {\mathsf E} \tau _ {ii} < \infty \ \textrm{ for any } i \in K , $$

where $ \tau _ {ii} $ is the return time to the state $ i $:

$$ \tau _ {ii} = \min \ \{ {t > 0 } : {\xi ( t) = i \mid \xi ( 0) = i } \} $$

for a discrete-time Markov chain, and

$$ \tau _ {ii} = \inf \ \{ {t > 0 } : {\xi ( t) = i \mid \xi ( 0) = i , \xi ( 0 + ) \neq i } \} $$

for a continuous-time Markov chain. When $ {\mathsf E} \tau _ {ii} = \infty $, $ K $ is called a zero class of states (class of zero states).

States in the same positive class $ K $ have a number of common properties. For example, in the case of discrete time, for any $ i , j \in K $ the limit relation

$$ \lim\limits _ {n \rightarrow \infty } \ \frac{1}{n} \sum _ { t= } 1 ^ { n } p _ {ij} ( t) = \ p _ {j} ^ {*} > 0 $$

holds; if

$$ d _ {i} = \max \ \{ {d } : { {\mathsf P} \{ \tau _ {ii} \ \textrm{ is divisible by } d \} = 1 } \} $$

is the period of state $ i $, then $ d _ {i} = d _ {j} $ for any $ i , j \in K $ and $ d $ is called the period of the class $ K $; for any $ i \in K $ the limit relation

$$ \lim\limits _ {t \rightarrow \infty } p _ {ii} ( t d ) = \ d p _ {i} ^ {*} > 0 $$

holds. A discrete-time Markov chain such that all its states form a single positive class of period 1 serves as an example of an ergodic Markov chain (cf. Markov chain, ergodic).

References

[1]	K.L. Chung, "Markov chains with stationary transition probabilities" , Springer (1967)
[2]	J.L. Doob, "Stochastic processes" , Wiley (1953)

Comments

Cf. also Markov chain, class of zero states of a for additional refences.