1
$\begingroup$

Let $H$ be a $\mathbb R$-Hilbert space, $\mu$ be a finite measure on $\mathcal B(H)$ with $\mu(\{0\})=0$ and $(L_t)_{t\ge0}$ be a $H$-valued càdlàg Lévy process on a probability space $(\Omega,\mathcal A,\operatorname P)$ with $$L_t=\sum_{i=1}^{N_t}Y_i\;\;\;\text{for all }t\ge0\tag1$$ for some $H$-valued independent identically distributed process $(Y_n)_{n\in\mathbb N}$ on $(\Omega,\mathcal A,\operatorname P)$ with $Y_1\sim\lambda^{-1}\mu$ for some $\lambda>0$ and some càdlàg Poisson process $(N_t)_{t\ge0}$ on $(\Omega,\mathcal A,\operatorname P)$.

Let $t\ge0$ and $B\in\mathcal B(H\setminus\{0\})$. I would like to show that $$\left|\left\{s\in[0,t]:\Delta L_s\in B\right\}\right|=\sum_{i=1}^{N_t}1_B(Y_i),\tag2$$ where $$\Delta L_s:=L_s-L_{s-}=L_s-\lim_{r\to s-}L_r\;\;\;\text{for }s\ge0.$$ How can we do that?

Remark: Since the measure is not involved in $(2)$ it might be unimportant for the claim, but we may note that $$Z_n:=\sum_{i=1}^nY_i\;\;\;\text{for }n\in\mathbb N$$ is a time-homogenous Markov chain and hence $$L_t=Z_{N_t}\;\;\;\text{for all }t\ge0$$ is a time-homogeneous Markov process.

$\endgroup$

1 Answer 1

1
$\begingroup$

$\newcommand{\De}{\Delta}$ This is not really a probability problem, since the equality \begin{equation}\label{1}\tag{1} l_t:=|\{s\in[0,t]\colon\De L_s\in B\}|=\sum_{i=1}^{N_t}1_B(Y_i)=:r_t \end{equation} holds almost everywhere on $\Omega$, for "almost" any joint distribution of the involved random variables -- provided only that $0\notin B$ and for all real $t\ge0$ \begin{equation*} N_t=\sum_{i=1}^\infty 1(\tau_i\le t), \tag{1.5} \end{equation*} where $0=\tau_0<\tau_1<\cdots$, so that for all natural $i$ and all real $t\ge0$ \begin{equation}\label{2}\tag{2} i\le N_t\iff\tau_i\le t. \end{equation} (Of course, here $\tau_1<\tau_2<\cdots$ are the times of the jumps of the Poisson process $N_\cdot$.)

Indeed, then for any $j\in\{0,1,\dots\}$ and any real $t\ge0$ \begin{equation*} L_t=\sum_{i=1}^{N_t}Y_i=\sum_{i=1}^j Y_i\quad\text{if}\quad \tau_j\le t<\tau_{j+1}. \end{equation*} Hence, for each real $s\ge0$, we have $\De L_s=Y_j$ if $s=\tau_j$ for some $j\in\{1,2,\dots\}$, and $\De L_s=0$ if $s\ne\tau_j$ for any $j\in\{1,2,\dots\}$. It follows that \begin{equation*} l_t=\sum_{j=1}^\infty 1(Y_j\in B,\tau_j\le t). \end{equation*}

On the other hand, for all real $t\ge0$ \begin{equation*} r_t=\sum_{i=1}^{N_t}1_B(Y_i) =\sum_{i=1}^\infty 1(Y_i\in B,i\le N_t)=\sum_{i=1}^\infty 1(Y_i\in B,\tau_i\le t), \end{equation*} by (\ref{2}).

Thus, $l_t=r_t$ for all real $t\ge0$, so that (\ref{1}) does hold.

$\endgroup$
4
  • $\begingroup$ You didn't specify your sequence $(\tau_k)_{k\in\mathbb N_0}$, but I guess $\tau_0:=0$ and $$\tau_k:=\inf\left\{t>\tau_{k-1}:\Delta N_t>0\right\}.$$ Since we know that there is a $\operatorname P$-null set $N$ such that $N(\omega)$ is nondecreasing and $$\Delta N_t(\omega)\in\{0,1\}\;\;\;\text{for all }t\ge0$$ for all $\omega\in\Omega\setminus N$, we can conclude that $$N_{\tau_k}(\omega)=k\;\;\;\text{for all }k\in\mathbb N_0\text{ and }\omega\in\Omega\setminus N.$$ This should yield your eq. $(2)$. $\endgroup$
    – 0xbadf00d
    Oct 23, 2020 at 4:44
  • $\begingroup$ @0xbadf00d : What you said in this comment is of course correct, but I don't understand the point of the comment. Is it to detail how (2) is obtained? In fact, (2) (as well as your description of the $\tau_i$'s) follow immediately from (1.5) and the condition $0=\tau_0<\tau_1<\cdots$. Anyway, other than this, are you satisfied with this answer? $\endgroup$ Oct 23, 2020 at 13:05
  • $\begingroup$ @0xbadf00d : To have a closure here, are you satisfied with my answer and the previous comment? $\endgroup$ Jan 8, 2021 at 17:08
  • $\begingroup$ I'll come back to this question and accept your answer soon. $\endgroup$
    – 0xbadf00d
    Jan 9, 2021 at 11:46

Your Answer

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

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