Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

A pure-birth Process is a generalization of a homogeneous Poission process. Whereas in the Poisson process the holding times between jumps are iid exponentially distributed random variables with parameter $\lambda$, in a pure-birth process the parameters may vary between jumps, i.e. the waiting between states $n$ and $n+1$ is exponentially distributed with parameter $\lambda_n$.

Let $X_1$, $X_1$ be two independent pure-birth processes with rates $\lambda^1_i, \lambda^2_i$ respectivley. By pairs of $X_1$ and $X_2$ I mean the process $$X(t) := (X_1(t), X_2(t))$$

I'd like to prove that $X$ is a contininuous-time Markov process that starts in state $(0,0)$ and may jump from state $(n,m)$ to either $(n+1,m)$ or $(n,m+1)$ with rates $\lambda^1_n$ and $\lambda^2_m$ respectively, depending on which process $X_1$,$X_2$ the next jumps occurs.

While this might seem very intuitive I'm having trouble relating the properties of $X_1$ and $X_2$ to $X$.

Can you point to a treatment I can apply my problem as a special case of?

share|improve this question
    
Got something from the answer below? –  Did Jul 29 '12 at 14:01
    
Yes, thank you very much! Thinking about it, of course it makes sense that it holds for Markov processes in general. I was able to find the same statement in Kallenberg's Foundation of Modern Probabilistics as an exercise. However i didn't fully get the notion of a infintesimal generator acting on functions. Is this another way to describe it instead of a matrix? And "acting" is just matrix multiplication? –  Haderlump Aug 2 '12 at 12:42
    
Indeed, to identify functions defined on the state space to vectors (of infinite size) is to identify the action of $L$ on functions to the multiplication (on the left) by a square-matrix (of infinite size). –  Did Aug 2 '12 at 13:34
add comment

1 Answer 1

up vote 1 down vote accepted

Every vector process whose coordinates are independent continuous-time Markov processes is a continuous-time Markov process. Here, $(X(t))$ is a continuous-time Markov process on the state space $S\times S$, where $S=\{0,1,2,\ldots\}$ is the common state space of the independent processes $(X_1(t))$ and $(X_2(t))$.

While in $(n,m)$, the process $(X(t))$ jumps at rate $\lambda_{n,m}=\lambda^1_n+\lambda^2_m$, that is, after an exponential time of mean $1/\lambda_{n,m}$. When this jump occurs, $(X(t))$ jumps to $(n+1,m)$ or to $(n,m+1)$ with respective probabilities $\lambda^1_n/\lambda_{n,m}$ and $\lambda^2_m/\lambda_{n,m}$ and independently of the time of the jump.

The infinitesimal generators $L_1$ and $L_2$ of $(X^1(t))$ and $(X^2(t))$ act on functions $v$ defined on $S$ by $$ L_iv(k)=\lambda^i_k\cdot(v(k+1)-v(k)),\qquad k\in S. $$ By definition, the infinitesimal generator $L$ of the independent product process $(X(t))$ acts on functions $u$ defined on $S\times S$ by $$ Lu(n,m)=L_1u_{\cdot,m}(n)+L_2u_{n,\cdot}(m),\qquad (n,m)\in S\times S, $$ where the functions $u_{n,\cdot}$ and $u_{\cdot,m}$ are defined on $S$ by $u_{n,\cdot}(m)=u_{\cdot,m}(n)=u(n,m)$. In the present case, one gets, for every $(n,m)$ in $S\times S$ $$ Lu(n,m)=\lambda^1_n\cdot(u(n+1,m)-u(n,m))+\lambda^2_m\cdot(u(n,m+1)-u(n,m)). $$

share|improve this answer
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.