2
$\begingroup$

Let's say there are independent waiting times of trains which are exponentially distributed with mean value $\frac{1}{2}$ hours. If you have already waited $1$ hour for the train, what's the probability that you will wait $2$ hours?

We have that mean value $\mu = \frac{1}{2}$. It's known that $\mu = \frac{1}{\lambda} \Leftrightarrow \frac{1}{2} = \frac{1}{\lambda} \Leftrightarrow \lambda = 2$

Now comes the part I'm not sure about.. when you have already waited $1$ hour before and now wait additional $2$ hours how the probability is calculated correctly here? I have looked this up on the internet and found that memorylessness applies to exponential distribution with formula:

$$P(X > r+t \mid X > r) = P(X>t)$$

where $r$ is the time you have previously waited, so $r=1$ and where $t$ is the time you have waited afterwards, thus $t=2$. Putting this into the formula we have

$$P(X>3 \mid X>1) = P(X>2) = 1-P(X<2) = 1-\left(1-e^{-2 \cdot2}\right) \approx 0.0183$$

So when you have already waited $1$ hour for the train, there is a probability of $0.0183$ that you will be waiting additional $2$ hours for the train?

Can you please tell me if it's correct like that because that's how I would do it in the exam :c

$\endgroup$
  • 2
    $\begingroup$ The word "additional" is not in the problem statement. You have to find $P(X>2\mid X>1)=P(X>1)=e^{-2\cdot 1}$. $\endgroup$ – NCh Feb 3 at 1:17
1
$\begingroup$

Let $T_n\stackrel{\mathrm{i.i.d.}}\sim\mathrm{Expo}(\lambda)$. Define $S_0=0$ and $S_n=\sum_{i=1}^n T_n$. Then $S_n$ is a renewal process, in fact a Poisson process, with associated counting process $N(t) = \sup\{n: S_n\leqslant t\}$. Define the age process by $A_t = t - S_{N(t)}$ and the residual process by $R_t = S_{N(t)+1} - t$ for $t\geqslant 0$. Define the renewal function $M(t) = \mathbb E[N(t)]$. Since $N(t)$ is Poisson distributed with mean $\lambda t$, it readily follows that $M(t)=\lambda t$. Let $Z_t\stackrel{\mathrm{def}}=(A_t,R_t)$. We derive the distribution of $Z_t$ by considering when $A_t=t$, that is, when $t$ lies within the first renewal interval. Then we have $$ f_{Z_t}(t,y) = f_{T_1}(t+y) = \lambda e^{-\lambda(t+y)}\cdot\mathsf 1_{[0,\infty)}(y). $$ The other case where $A_t<t$ corresponds to when $t$ occurs after the first renewal. We have \begin{align} f_{Z_t}(x,y) &= \sum_{n=1}^\infty \frac{\lambda^n(t-x)^{n-1}e^{-\lambda(t-x)}}{(n-1)!}\lambda e^{-\lambda(x+y)}\ &= \lambda^2 e^{-\lambda(x+y)}\cdot\mathsf 1_{[0,t)\times[0,\infty)}(x,y).

Combining these two cases yields

$$ f_{Z_t}(x,y) = \lambda^2e^{-\lambda(x+y)}\mathsf 1_{[0,t)\times[0,\infty)}(x,y) + \lambda^2e^{-\lambda(t+y)}\mathsf 1_{\{x=t\}\times[0,\infty)}(x,y) $$

| cite | improve this answer | |
$\endgroup$

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.