I got this problem where:

The random variables $X_1, X_2,$ and $X_3$ are independent and $Exp(1)-$ distributed.
Compute the correlation coefficient $r(X_{(1)},X_{(3)})$ .

I know through research that: $\mathrm E(X_{(k)})=\sum\limits_{i=n-k+1}^n\frac1i,\qquad \mbox{Var}(X_{(k)})=\sum\limits_{i=n-k+1}^n\frac1{i^2}.$

I could find everything except, $E(X_{(1)}X_{(3)})$.

I got: $E(X_{(1)})=\frac {1}{3},\qquad E(X_{(3)})=1+\frac {1}{2}+\frac {1}{3}$

$Var(X_{(1)})= \frac {1}{9}, \qquad Var(X_{(3)})= 1+\frac {1}{4}+\frac {1}{9}$

I also know that correlation = $$ \frac{cov(X_{(1)},X_{(3)})}{\sqrt{Var(X_{(1)})Var(X_{(3)})}}$$


So from the implication of @NcH I figured the $E(X_{(1)}X_{(3)})=1$ But, when trying to compute the correlation coefficient I get 1 as the final answer as opposed to 2/7 suggested by the textbook.

$$r= \frac{1-11/18}{7/18}=1$$

Anyone has a different answer ?

  • 1
    $\begingroup$ Your solutions assume that the sample size is 3, but this not stated in the question. The answer should therefore be a function of the sample size $n$. $\endgroup$ – wolfies Feb 21 at 12:21
  • $\begingroup$ Side note, the answer is 2/7 $\endgroup$ – Mahamad A. Kanouté Feb 21 at 12:23
  • $\begingroup$ @wolfies it is explicitly stated in the question, while OP used the results which are true for sample size $n$ and $k$-th order statistics. $\endgroup$ – pointguard0 Feb 21 at 12:35
  • $\begingroup$ @pointguard Why would the expectation of the sample minimum in a sample of size $n$ be equal to $\frac13$? $\endgroup$ – wolfies Feb 21 at 12:40
  • $\begingroup$ it is not equal to $\frac 1 3$, it is equal to $\sum_{i=n-k+1}^n \frac 1 i$. $\endgroup$ – pointguard0 Feb 21 at 12:41

I think that the simplest way is to use joint PDF of $X_{(1)}$ and $X_{(3)}$. For $0<x<y$ using formula for [joint pdf of two order ststistics]1 we get $$ f_{X_{(1)}, X_{(3)}}(x,y) = 3! \bigl((1-e^{-y})-(1-e^{-x})\bigr)e^{-x}e^{-y}=6(e^{-2x-y}-e^{-x-2y}) $$

Then $$\mathbb E[X_{(1)}X_{(3)}]=\iint_{0<x<y<\infty} xy f_{X_{(1)}, X_{(3)}}(x,y) \,dx\, dy.$$

  • $\begingroup$ I guess you're just missing the 6 at the end of your joint. Aside that thank you for your reply $\endgroup$ – Mahamad A. Kanouté Feb 21 at 23:53
  • $\begingroup$ I got 1 as final answer even though the solution suggest 2/7 $\endgroup$ – Mahamad A. Kanouté Feb 22 at 0:43
  • $\begingroup$ Thanks, I added lost $6$. No, this integral is not equal to $1$. Wolframalpha gives answer $\frac{13}{18}$. $\endgroup$ – NCh Feb 22 at 2:42
  • $\begingroup$ x goes from 0 to y and y goes from 0 to $\infty$ right ? $\endgroup$ – Mahamad A. Kanouté Feb 22 at 2:45
  • $\begingroup$ Oh yeah you're right I see my error $\endgroup$ – Mahamad A. Kanouté Feb 22 at 2:46

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.