Waiting for a train in a subway

  • Thread starter Thread starter docnet
  • Start date Start date
  • Tags Tags
    Train
AI Thread Summary
The discussion revolves around the calculation of waiting times for a train using exponential probability distributions. It highlights the integration of functions to determine probabilities and expected values, specifically addressing the issues of normalization and the choice of distribution functions. Participants debate the appropriateness of different probability distributions, noting that the first option does not satisfy the criteria for a probability distribution, while the second is normalized but may not align with the expected exponential behavior. The conversation also touches on the implications of introducing a new variable to represent remaining time and the consistency of results across different calculations. Overall, the analysis emphasizes the importance of correctly interpreting and applying exponential distributions in real-world scenarios.
docnet
Messages
796
Reaction score
488
Homework Statement
.
Relevant Equations
.
Screen Shot 2021-11-29 at 1.34.24 AM.png

(a) $$P(X≥10)=∫_{10}^∞10e^{−10x}dx=e^{−100}$$

(b) $$E(X)=∫_0^∞10xe^{−10x}dx=\frac{1}{10}$$

(c)$$\frac{10e^{−10}}{P(X≥10)}=10e^{100−10x}=\text{the new probability distribution for}\quad x≥10$$
$$E(X)=∫_{10}^∞10xe^{100−10x}dx=\frac{101}{10}$$😒

(d) imho it's unclear which "probability distribution functions" I should use here.
$$\begin{cases}
10e^{-10x} & 0\leq a\leq 10\\
0 & \text{otherwise}
\end{cases}$$
$$\begin{cases}
\frac{10e^{-10x} }{1-e^{-100}} & 0\leq a\leq 10\\
0 & \text{otherwise}
\end{cases}$$
The first doesn't satisfy the definition of the probability distribution function because its integral over ##\mathbb{R}## is less than 1. The second one has been normalized to 1 but it is not equal to ##X\sim \exp⁡(10)##. The first leads to ##\frac{1-101e^{-100}}{10}## minutes. The second leads to ##\frac{1}{10(1−e^{−100})}## minutes which does not parse with the fact that train always arrives by 10 minutes at the latest, so it should be less than 110 minutes.
 
Last edited:
Physics news on Phys.org
First, at this level I would prefer to remain with the generic exponential distribution, as we can see more clearly what's going on. Just plug in ##\lambda = 10## and ##t_0 = 10## to get the numerical answers.

For example, for part a) we see that $$P(X \ge t_0) = e^{-\lambda t_0} = e^{-100}$$

For part c) I suggest you need to introduce a new variable ##X' = X - t_0##, which represents the remaining time beyond ##t_0##, and ##X## represents the total time including ##t_0##.

I don't agree with your answer, because you want ##E(X')## instead of ##E(X)##.

For part d), conceptually we have a reduced sample space of those cases where ##X \le t_0##. We may, therefore, take the truncated distribution for ##X \le t_0## and scale it up by the relevant factor. In this case:
$$f(x) = \frac{\lambda e^{-\lambda x} } {\int_0^{t_0}\lambda e^{-\lambda x}dx} \ (0 \le x \le t_0)$$Which is equivalent to your second option.
 
  • Love
Likes docnet
Are you sure that by exp(10) they mean something that looks like ##e^{-10 x}## and not something like ##e^{-x/10}##?

The question in general makes a lot more sense if it's the latter. Perhaps a typo in the book?
 
  • Like
Likes docnet and PeroK
PeroK said:
For part c) I suggest you need to introduce a new variable ##X' = X - t_0##, which represents the remaining time beyond ##t_0##, and ##X## represents the total time including ##t_0##.
thank you :)

$$X'=X-t_0$$
$$E(X')=E(X)-E(t_0)=\frac{101}{10}-10=\frac{1}{10}$$
Do you think it strange the answers for (b) and (c) are same?

Office_Shredder said:
Are you sure that by exp(10) they mean something that looks like ##e^{-10 x}## and not something like ##e^{-x/10}##?

The question in general makes a lot more sense if it's the latter. Perhaps a typo in the book?
yes, I'm sure the professor means ##\lambda e^{-\lambda x}##
Screen Shot 2021-11-29 at 2.09.10 PM.png
 
docnet said:
Do you think it strange the answers for (b) and (c) are same?
No, not strange. Inevitable, if you do the maths. Once an exponential, always an exponential!
 
And nobody is objecting ? If the waiting time is really six seconds, the train can't even stop ! :oldlaugh:

##\ ##
 
BvU said:
And nobody is objecting ? If the waiting time is really six seconds, the train can't even stop ! :oldlaugh:

##\ ##
It's called a bullet train!
 
  • Haha
Likes BvU
PeroK said:
No, not strange. Inevitable, if you do the maths. Once an exponential, always an exponential!
To see this, consider the distribution for ##x' = x - t_0## in the cases where we have waited a time ##t_0## without a train. We have:
$$P(X \ge t_0) = \int_{t_0}^{\infty} \lambda e^{-\lambda x} dx = e^{-\lambda t_0}$$That's the size of the new sample space. And:
$$f(x') = \frac{\lambda e^{-\lambda(x' + t_0)}}{e^{-\lambda t_0}} = \lambda e^{-\lambda x'}$$Which shows that the distribution for the remaining time remains the same exponential distribution.

That's a good example of what you discover when you stick to the general case and resist the temptation to plug and chug!
 
  • Love
Likes docnet
Back
Top