We have a sample $X_1, ..., X_n$ from $U[M,2M]$, $M>0$. Define $U=\frac{(n+1)}{(2n+1)}\max(X_i)$. I wish to show that $U$ is an unbiased estimator for $M$, but I am getting a wrong result. This is how I am trying to do it.


$P(X_i \leq x)=\int_M^x \frac{1}{M} dy = \frac{x-M}{M}$

$P(\max(X_i) \leq x) = P(X_1 \leq x, ... , X_n \leq x)=(\frac{x-M}{M})^n$

$E(\max(X_i))=\int_M^{2M} {1- (\frac{y-M}{M})^n } dy = \frac{n}{n+1}M$


but I have to show that $ E(U)=M$ so it is an unbiased estimator, I really don't see where I am going wrong!

  • $\begingroup$ When computing E(max(X_i)) you forgot to integrate P(max(X_i)>x)=1 from x=0 to x=M. $\endgroup$ – Did Jun 8 '14 at 20:39
  • $\begingroup$ Thank you, that was indeed the mistake. $\endgroup$ – Shawn Jun 8 '14 at 20:54

It looks like your problems start occurring when you try to calculate the Expected Value of X.

Go back one step: When you calculate the probability that all X's are less than "x", you're in effect calculating the CDF, right? As a result, you want to take the derivative of this CDF, giving you your new PDF to work with [i.e. n/M(X/M -1)^(n-1)]. [See: Expected Value of Max of IID Variables ]

The next step would be to multiply by X and integrate from M to 2M (performing the necessary u-substitution, etc.), leaving you with the Expected Value of X.

From there it's easy since the "2n+1" and "n+1" cancel out the "2n+1" and "n+1" earlier...

| cite | improve this answer | |

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.