Perron- Frobenius for a Continuous time Markov Chain

411 Views Asked by At

I'm reading a paper by Villemonais and Mèlèard , where it is used the Perron Frobenius theorem for a continuous in time Mkv Chain. It is stated that there exists a positive eigenvalue for the matrix $ P_{t_0}$ (where $ (P_{t})$ is a sub-Markovian semi-group of probabilities) with $t_0$ such that $P_{t_0}$ has only positive entries.

But I had read Darroch and Seneta (1967) where the theorem is stated for $Q$ the infinitesimal generator.

What is the relationship between both versions? Are both correct?

Where can I find the theorem for continuous time properly stated (and proved if possible)?

How does this statement change when considering a Mkv Chain on an infinite state space?

1

There are 1 best solutions below

2
On

The classical version of Perron Frobenius theorem is stated for positive matrices (as in the MV paper you're citing). See for instance the wikipedia entry https://en.wikipedia.org/wiki/Perron%E2%80%93Frobenius_theorem#Positive_matrices. There exist more powerful versions of the Perron-Frobenius theorem, but this one should cover your need.

I don't find any reference to an infinitesimal generator in Darroch and Seneta 1965. However, in general, if you have a Perron-Frobenius theorem for an infinitesimal generator $Q$, then you immediately obtain the same kind of conclusion for $P_{t_0}$ through the equality $P_{t_0}=e^{Qt_0}$.