4
$\begingroup$

I'm sorry if this question is trivial but I'm new at this and I'm learning this stuff on my own from a book + youtube tutorials so I haven't really got anyone to ask. I have the following exercise from a book of basic probability theory:

"A process is divided into 3 sub-processes which are carried out in parallel. Each sub-process takes an Exponential amount of time, 5 minutes on the average, independent of the other sub-processes. The process is finished when all sub-processes are finished. What is the expected time for the full process to complete?"

If I understand this correctly we have three random variables $X_1, X_2, X_3 \sim Exponential(1/5)$. My naive guess would be that the expected time is 5 minutes as the sub-processes are independent and parallel but I'm probably missing something. I guess I could compute the joint distribution $f(Y) = f(X_1, X_2, X_3)$ and calculate $E(Y)$ from this. However, this entails calculating a triple integral and I'm hoping something in the simple setup (independence and identical distributions) means there is a simpler solution. Can anyone give me any advice?

$\endgroup$
5
  • 1
    $\begingroup$ +1 Begin by writing down a formula for the time to completion based on the actual times needed by the three subprocesses. For instance, when those times are $1$, $8$, and $9$ minutes, your formula should give $9$ minutes. This formula, when applied to $X_1,X_2,X_3$, gives a univariate random variable. Find its distribution function (it doesn't require a triple integral), then compute the expectation. $\endgroup$
    – whuber
    Sep 15, 2015 at 19:54
  • $\begingroup$ As Whuber suggested, "the process is finished when all sub-processes are finished" leads to a mathematical description of one variable of interest, depending on these three random variables. $\endgroup$ Sep 15, 2015 at 20:33
  • $\begingroup$ @whuber I'm sorry but I'm not sure I understand. Are you saying I should construct a formula f(X1,X2,X3) = max(X1,X2,X3) and calculate it's density? $\endgroup$ Sep 16, 2015 at 12:38
  • 1
    $\begingroup$ @whuber Okay, I think I got it. P(Y<t) = P(X1<t, X2<t, X3<t)=/by independence/=P(X1<t)P(X2<t)P(X3<t). The CDF of Y would then be (1-e^(-t/5))^3. I derive this to get PDF, f(t), and calculate E(Y) the usual way by integrating t*f(t) to infinity. This gave me ~9.17 which doesn't seem unreasonable. Is my reasoning correct? $\endgroup$ Sep 16, 2015 at 15:14
  • 1
    $\begingroup$ Sounds right to me. A good way to get more feedback is to post your solution as an answer, then wait for votes and comments. Incidentally, there's an easier way to obtain the expectation: integrate $1-F(x)^3$ from $0$ to $\infty$ (where $F$ is the common CDF of the $X_i$). $\endgroup$
    – whuber
    Sep 16, 2015 at 15:52

1 Answer 1

2
$\begingroup$

Okay, I think I got it. $P(Y<t) = P(X_1<t, X_2<t, X_3<t) = /\text{by independence}/ = P(X_1<t)P(X_2<t)P(X_3<t)$

The CDF of $Y$ would then be $(1-e^{-t/5})^3$. I derive this to get PDF, $f(t)$, and calculate $E(Y)$ the usual way by integrating $t \times f(t)$ to infinity. This gave me ~9.17 which doesn't seem unreasonable. Is my reasoning correct?

$\endgroup$

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.