Exercise 1 (Exponential random random variables) Let \(E_1, \ldots, E_n\) be independent exponential random variables with rates \(\lambda_1, \ldots, \lambda_n\) Show that \[\mathbb{P} ( E_1 = \min (E_1, \ldots, E_n)) = \frac{\lambda_1}{ \lambda_1 + \cdots + \lambda_n}.\]
Exercise 2 (Q matrices) Consider the continuous-time Markov chain \(X\) with three state \(\{1,2,3\}\) with \(Q\) matrix given by
Q <- matrix(c(-6,3,3, 2,-3,1, 2,7,-9), nrow =3)
Q = t(Q)
Q
## [,1] [,2] [,3]
## [1,] -6 3 3
## [2,] 2 -3 1
## [3,] 2 7 -9
- Find the stationary distribution \(\pi\).
- Start the Markov chain at state \(i\), and write code so that you obtain the state of the chain at time \(t\).
- Starting at state \(1\), see what state the chain is in after a large time \(t\); repeat for a large number of times.
- On average, how often is \(X(t) =3\) for large \(t\)?
- Discuss this experiment in relation to the theory we discussed.
Exercise 3 (Stationary measures) Let \(P\) be a transition matrix semigroup for an irreducible continuous-time Markov chain on a finite number of states \(A\) with the stationary measure \(\pi\). Let \(Q\) be the generator. Let \(M\) be transition matrix for the corresponding jump chain with the corresponding stationary measure \(\hat{\pi}\). Show that
\[\hat{\pi}_i = \frac{q_{ii} \pi_i}{\sum_j q_{jj} \pi_j }.\]