Odd event in counting process

241 Views Asked by At

Let's say $\{X_1,X_2,X_3,...\}$ is a set of positive IID random variable has distribution on a general distribution F

The summation of these random variable is a renewal sequence

$$S_n=X_1+X_2+X_3+...+X_n$$

Let N(t) define as a counting process that $ N(t)=\sum_{n=1}^{\infty}\mathbf{1}_{(0,t]}(S_n)$.

We would like to find the probability that $P[N(t) \text{ is odd}]$

My reasoning for this problem is define a new $N'(t)=\sum_{n=1}^{\infty} \mathbf{1}_{(0,t]}(S_{2n-1})$ but I have no idea how to derive the probability out of $N'(t)$ since the distribution of $X$ is general. My guess is that $P(N'(t))$ will be some kind of renewal type equation.

This is a question that I think is a general version for the splitting event of Poisson process discussed before.

Update 1

I tried to borrow the idea from Poisson process splitting problem. The interarrival time for the new counting process $N'(t)$ is $X_i+X_{i+1}$ which is the convolution of the distribution $F(t)*F(t)=F^{2*}$. If we know about the new interarrival distribution, can we derive the probability out of it?

Update 2

I think I come up a solution but I would appriciate if others could check my answer.

For a renewal process $$N(t) \ge n \Leftrightarrow S_n \leq t$$ (I do confuse about this relation but it appears in many stochastic process textbooks)

Therefore

$$ \begin{align} P[N(t) = n] &= P[N(t)\ge n]-P[N(t) \ge n+1] \\ &=P[S_n \le t]-P[S_{n+1} \le t] \end{align} $$

Thus

$$ \begin{align} P(N(t) \text{ is odd}) &= \sum_{n=1}^{\infty}\{P(S_{2n-1}\le t) - P(S_{2n}\le t)\} \\ &=\sum_{n=1}^{\infty}(F^{(2n-1)*}-F^{(2n)*}) \end{align} $$

Here, $F^{n*}$ is n-fold convolution for the distribution F (from sum of random r.v)