FIND ME ON

GitHub

LinkedIn

Kac's Lemma

🌱

Definition
StochasticProcesses

Let {xt}\{ x_{t} \} be irreducible and π\pi its Invariant probability measure. Then it is unique and fix kXk\in \mathbb{X} πi=γikEk[Tk(1)]πk=γkkEk[Tk(1)]=1Ek[Tk(1)]\begin{gather*} \pi_{i}=\frac{\gamma_{i}^{k}}{E_{k}[T_{k}^{(1)}]}\\\\ \pi_{k}=\frac{\gamma_{k}^{k}}{E_{k}[T_{k}^{(1)}]}=\frac{1}{E_{k}[T_{k}^{(1)}]} \end{gather*}

If you know π\pi then mk=Ek[Tk(1)]=1πkγik=πiπk\begin{align*} m_{k}=E_{k}[T_{k}^{(1)}]=\frac{1}{\pi_{k}}\\\\ \gamma_{i}^{k}=\frac{\pi_{i}}{\pi_{k}} \end{align*} In a sense, γik\gamma_{i}^{k}​ reflects the proportion of time the chain spends in state ii relative to state kk in the long run, which is a key concept in determining the invariant distribution.

Linked from