I encounter a problem in which I need to characterize the asymptotic behaviour of a sequence.

$\{s_{n,k}\}$ is a stationary probability distribution, i.e., $\sum_{n=0}^{\infty}{\sum_{k=0}^{\infty}{s_{n,k}}=1}$, and satisfies
$s_{n,k}=0$ for $n<0$ or $k<0$,
and $(n\mu+k\beta+\delta)s_{n,k}=(k+1)\beta s_{n-1,k+1}+n\mu s_{n,k-1}$,
where $\alpha, \delta, \beta, \mu$ are constants.

What I want to know is the asymptotic behaviour of $\sum_{n=m}^{\infty}{\sum_{k=0}^{\infty}{s_{n,k}}}$ when $m \rightarrow \infty$

Initially I considered the generating function $F(x,y)=\sum_{n=0}^{\infty}{\sum_{k=0}^{\infty}{s_{n,k}x^ny^k}}$ which satisfies the PDE $\delta F=\beta(x-y)F_y+\mu x(y-1)F_x+\alpha y$, but I didn't find anything useful since this PDE doesn't seem solvable.

Can anyone suggest a way to do it or suggest references? This problem arises in an economic model for firms (you can think of $n$ as size and $k$ as the number of vacancies, and thus the recursive relation describes how firm size grows). It's like a generalization of the birth-death process, but is more complicated.

Thank you for your help!

  • $\begingroup$ Maybe a obvious remark. You have to impose boundary values for $(s_{0,k})$ with $k=0,\dots,\infty$, not only for $k=1$. The reason is, that for $n=0$ your recurrence relation is meaningless as it would impose $s_{0,k}=0$. Then, a question is how fast has to decay the sequence, such that $(s_{n,k})$ is normalizable to a probability distribution? $\endgroup$ – André Schlichting Jul 27 '11 at 8:31
  • $\begingroup$ Actually $s_{0,k}=0$ for $k\neq 1$ is the boundary condition I want. $\endgroup$ – Epsilon Jul 27 '11 at 18:18

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.