2
$\begingroup$

In a Markov chain, we have four states: $0, 1, 2, 3$. The transition matrix is $$ \begin{bmatrix} 1/3 & 2/3 & 0 & 0 \\ 1/3 & 0 & 2/3 & 0 \\ 0 & 1/3 & 0 & 2/3\\ 0& 0 & 0 & 1 \end{bmatrix}$$ The initial state is $1$, and we need to find the probability that $0$ is visited at least twice before reaching $3$ (which is "game over": an absorbing state).

My solution if the following: Define $P_{i v}$ as the probability to visit $0$ at least $v$ times before reaching $3$, given that the current state is $i$ (not including $i$ if it happends to be $0$).

Then we can use the law of total probability to obtain: $$ {\left\{\begin{matrix} P_{12}=\frac{2}{3}P_{22}+\frac{1}{3}P_{01}\\ P_{22}=\frac{2}{3}\cdot 0+\frac{1}{3}P_{12}\\ P_{01}=\frac{2}{3}P_{11}+\frac{1}{3}\cdot 1\\ P_{11}=\frac{2}{3}P_{21}+\frac{1}{3}\cdot 1\\ P_{21}=\frac{2}{3}\cdot 0+\frac{1}{3}P_{11}\\ \end{matrix}\right.} $$ Which has the solution: $$ {\left\{\begin{matrix} P_{12}=13/49\\ P_{22}=13/147\\ P_{01}=13/21\\ P_{11}=3/7\\ P_{21}=1/7\\ \end{matrix}\right.} $$ so the answer seems to be $13/49$. However, the solution attached (which seems to never be wrong) says that the answer is $95/243$. So I have two questions:

  1. Could someone please verify my solution?
  2. Does someone know a shorter solution?

Thanks!

  • 0
    Doing a computation by hand, I also get a solution $13/49$.2017-02-12
  • 0
    You mean by solving the equation system?2017-02-12
  • 0
    My approach was: (1) eliminate state $2$, and say that we can go from $1$ to $3$ with probability $4/9$ and from $1$ to $1$ with probability $2/9$; (2) eliminate the self-loop in state $1$, so that conditioned on leaving state $1$, you either go to $0$ (with probability $3/7$) or to $3$ (with probability $4/7$); (3) state $3$ finally "wins" with probability $4/7 + 3/7 \cdot 2/3 \cdot 4/7$ (directly or one visit to $0$). Finally $1 - (4/7 + 3/7 \cdot 2/3 \cdot 4/7) = 13/49$.2017-02-12
  • 0
    Nice method! Thank you, that helped2017-02-12

0 Answers 0