The observations $X_1,\dots, X_n$ form an IID sample from a probability distribution with unknown mean $U$. Let

$$ \hat{U}=\frac{X_2+X_3+\dots+X_n}{n+10}$$

be an estimator used to estimate $U$. (I.e., compared to the more usual estimator $\bar{X}$, $\hat{U}$ discards the first observation and uses the divisor $n+10$ rather than $n$). The question I want to answer is what is the bias of $\hat{U}$?

I am aware that the bias ($b$) of an estimator ($\hat{U}$) for $U$ is

$$ b=E(\hat{U})-U $$,

that is, the bias is equal to the expectation of the estimator minus the actual value of the estimand. In this case

$$ b=E\bigg(\frac{X_2+X_3+\dots+X_n}{n+10}\bigg)-U.$$

I know that I cannot get a specific value but the question I have been given has asked me to simplify this question down and I am unsure how.

Also in an example given it states that the bias of $\hat{U}$ is $-U/(N+1)$, but I am unclear how to get this answer.

  • $\begingroup$ What are you trying to estimate? In your formula for $b$ and in the title you begin indexing with $2$: is that on purpose or is it a typographical error? $\endgroup$ – whuber Jan 4 '18 at 14:51
  • $\begingroup$ Can you clarify what U represents in this problem? Also, do we know the expected value of the $X_i$s? Do they come from the same distribution? $\endgroup$ – Anshu says Reinstate Monica Jan 4 '18 at 14:54
  • 1
    $\begingroup$ Thank you. When you incorporate that information in your post, it will come to the attention of the community an enable people to vote to reopen it. $\endgroup$ – whuber Jan 4 '18 at 17:04
  • 1
    $\begingroup$ Thank you for the edit. I have tried to improve the post a little more and voted to reopen. I also added the self-study tag and would recommend that you read its tag wiki. Here is a first hint to get you started: you need to get the expectation of $\hat{U}$. How can you calculate it, given that the $X_i$ are IID? $\endgroup$ – Stephan Kolassa Jan 4 '18 at 18:43
  • 1
    $\begingroup$ the bias given in the example you cite $b=-U/(n+1)$ is wrong. There must be a couple typing mistakes in your book. $\endgroup$ – DeltaIV Jan 5 '18 at 10:20

I'm not sure how much detail to give in an answer to a self-study question, but I hope that this will help you get started:

The trickiest part is getting $b$ in terms of $U$ and $n$ so that we can express it as a function of parameters, rather than the random variables $\{{X_i}\}^{n}_{i=2})$. It helps to observe that expectation is linear. That is, $E(X_{1}+X_{2})=E(X_1)+E(X_2)$. Try applying this rule to the equation for $b$:



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.