3
$\begingroup$

Let $X_1,...,X_n$ be a random sample of size $n$ from a Bernoulli distribution with parameter $p$ where $0< p< 1$ is unkown. (a) Find $\theta^2=Var(\bar{X}).$ (b) Find the value of $c$ so that $c\bar{X}(1-\bar{X})$ is an unbiased estimator of $\theta^2$.

My attempt:

(a) Since we are sampling from a Bernoulli distribution, $\sum_{i=1}^n X_i$~$Binomial(n,p)$. Hence, $\mu_{\bar{X}}=\frac{1}{n}\mu_{\sum Xi}=\frac{1}{n}np=p,$ and

$Var(\bar{X})=\frac{1}{n^2}Var(\sum X_i)=\frac{1}{n^2}np(1-p)=\frac{p(1-p)}{n}$

(b)$E(\bar{X}^2)=V(\bar{X})+\mu^2_{\bar{X}}=\frac{p(1-p)}{n}+p^2$. Hence, $E(c\bar{X}(1-\bar{X})=c[E(\bar{X})-E(\bar{X}^2)]=c[p-(\frac{p(1-p)}{n}+p^2)]=c(\frac{np-p+p^2-np^2}{n}).$ To be unbiased, this must equal $Var(\bar{X})$. Hence $c(\frac{np-p+p^2-np^2}{n})=\frac{p(1-p)}{n}\Rightarrow c=\frac{1-p}{n-1+p-np}.$

Does this look ok? My answer for part (b) doesn't seem right to me

$\endgroup$
  • 1
    $\begingroup$ Please add the self-study tag to these sorts of questions. $\endgroup$ – Glen_b Oct 7 '14 at 15:18
  • 1
    $\begingroup$ As a quick reality check, in the case $n=1$ obviously $X=\bar X=(\bar X)^2$ (since all values are either $0$ or $1$), whence $\mathbb{E}(\bar X(1-\bar X))=0$. Indeed, that's what $(np-p+p^2-np^2)/n=(n-1)/n$ is equal to, so you're doing something right in (b). $\endgroup$ – whuber Oct 7 '14 at 16:50
3
$\begingroup$

Don't know if you made any mistakes along the way but you could simplify your final result, as long as $p \neq 1$.

$$\frac{1-p}{n-1+p-np} = \frac{ (1-p)}{(n-1)(1-p)} = \frac{1}{n-1} $$

$\endgroup$
4
$\begingroup$

You might want to try to simplify $np-p+p^2-np^2$. Maybe factor out $n-1$ and $1-p$. I didn't check your final answer, but it might be correct, just less-pretty. As a hint $c$ should not be a function of $p$.

$\endgroup$

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.